UCGE Reports

Number 20184

Department of Geomatics Engineering

IF GPS Signal Simulator Development and Verification

(URL: http://www.geomatics.ucalgary.ca/links/GradTheses.html)

by

Lei Dong

December 2003
 THE UNIVERSITY OF CALGARY

IF GPS Signal Simulator Development and Verification

by

Lei Dong

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF SCIENCE

DEPARTMENT OF GEOMATICS ENGINEERING

CALGARY, ALBERTA

November, 2003

© Lei Dong 2003

 ABSTRACT

A software-based intermediate frequency (IF) GPS signal simulator for the L1 C/A code

signal is developed and verified in this thesis. The signal simulator is implemented in

MATLAB with some key functions coded in C to accelerate speed of operation. A

mathematical signal model is developed that expresses the digitized IF GPS signal as a

function of various errors during propagation, such as satellite clock error, ionosphere

error, and troposphere error; and user defined parameters, such as date and GPS time,

user trajectory, expected C/N0, noise density, intermediate frequency, sampling rate,

front-end bandwidth, and quantization bit. The simulator is verified by comparing the

simulated signal with the hardware-collected IF GPS signal in the time, frequency,

correlation and position domains.

Partially for the simulator verification purpose, a software GPS receiver was developed

and verified, which was used to process the simulated signal to get the pseudorange and

carrier phase measurements. The signal acquisition and tracking techniques originally

studied and verified in this thesis were used for the first generation software receiver

developed by the PLAN group of the Department of Geomatics Engineering. The

receiver consists of the following modules: signal acquisition, signal tracking, bit

synchronization, sub-frame synchronization and the navigation solution. It is verified by

using a set of real GPS signal measurements collected by a commercial hardware front-

end and comparing the tracking results and navigation solution with the corresponding

results of a hardware GPS receiver.

iii
 ACKNOWLEDGEMENTS

I would like to express my sincere appreciation to my supervisor, Professor Gérard

Lachapelle, for his continuous guidance and support during my Master’s program. I will

always remember his sincere spirit of cooperation, his positive attitude, and

understanding.

I would like to acknowledge the following graduate students and research associates:

Olivier Julien, Aaron Morton, Changlin Ma, and Mark Petovello for their support and

kindness.

Thanks also to the professors, students, and staff of the Department of Geomatics

Engineering who have made my time fruitful and enjoyable.

My deepest thanks go to my dear husband, Changlin Ma, who supported me

tremendously throughout my study, and to our son, Xingyu, for bringing so much joy into

my life. I am also indebted to my parents and parents-in-law for their untiring support.

iv
 TABLE OF CONTENTS

ABSTRACT............................................................................................................................ iii
ACKNOWLEDGEMENTS .................................................................................................. iv
TABLE OF CONTENTS ....................................................................................................... v
LIST OF TABLES ................................................................................................................. ix
LIST OF FIGURES ................................................................................................................ x
NOTATION.......................................................................................................................... xvi
0.1 Symbols ........................................................................................................................... xvi
0.2 Abbreviations and Acronyms ....................................................................................... xix

CHAPTER 1

INTRODUCTION................................................................................................................... 1
1.1 Background ....................................................................................................................... 1
1.2 GPS Signal Simulation Review ........................................................................................ 3
1.3 Objective ............................................................................................................................ 6
1.4 Thesis Outline.................................................................................................................... 7

CHAPTER 2

GPS SIGNAL AND RECEIVER THEORY ........................................................................ 9

2.1 GPS Signals ....................................................................................................................... 9
2.1.1 GPS Signal Structure ....................................................................................................9
2.1.2 GPS Signal Power Levels ...........................................................................................12
2.1.3 L1 C/A Code Signal in the Frequency Domain..........................................................14
2.1.4 C/A Code Auto-correlation and Cross-correlation .....................................................16
2.2 Conventional Hardware Front-end and GPS Receiver............................................... 17
2.2.1 Front-end.....................................................................................................................19

v
 2.2.2 Signal Processing........................................................................................................21
2.3 Error Sources .................................................................................................................. 25
2.3.1 Ephemeris Errors ........................................................................................................26
2.3.2 Satellite Clock Error ...................................................................................................26
2.3.3 Ionospheric Error and Tropospheric Error..................................................................27
2.3.4 Multipath.....................................................................................................................28
2.3.5 Receiver Noise Errors.................................................................................................29

CHAPTER 3

IF GPS SIGNAL SIMULATOR REALIZATION ............................................................ 30

3.1 IF GPS Signal Simulation Model Derivation ............................................................... 30
3.1.1 Transmitted GPS Signal..............................................................................................31
3.1.2 Propagation through the Atmosphere to the Antenna.................................................34
3.1.3 Down-conversion........................................................................................................37
3.1.4 Intermediate Frequency Processing ............................................................................39
3.2 Simulator Realization ..................................................................................................... 39
3.2.1 Satellites in View ........................................................................................................41
3.2.2 Amplitude Modeling...................................................................................................43
3.2.3 Navigation Data Simulation........................................................................................44
3.2.4 C/A Code Modeling....................................................................................................46
3.2.5 Signal Propagation Time and Doppler........................................................................47
3.2.6 Satellite Clock Error Modeling...................................................................................51
3.2.7 Ionospheric Error Modeling........................................................................................52
3.2.8 Tropospheric Error Modeling .....................................................................................59
3.2.9 Noise ...........................................................................................................................63
3.2.10 Filtering.....................................................................................................................65
3.2.11 Quantization..............................................................................................................67
3.2.12 Program Flow Chart..................................................................................................71

vi
CHAPTER 4

SOFTWARE GPS RECEIVER ......................................................................76

4.1 Introduction.................................................................................................76
4.2 Signal Acquisition ........................................................................................................... 77
4.2.1 DFT-based Acquisition...............................................................................................79
4.3 Signal Tracking ............................................................................................................... 82
4.3.1 Basic Phase Lock Loop...............................................................................................83
4.3.2 Delay Lock Loop ........................................................................................................86
4.3.3 Carrier Tracking Loop ................................................................................................90
4.4 DLL Lock Detectors and C/N0 Estimation ................................................................... 94
4.5 Bit Synchronization ........................................................................................................ 96
4.6 Sub-frame Synchronization ........................................................................................... 98
4.7 Raw Measurement Derivation ..................................................................................... 102
4.7.1 Pseudorange ..............................................................................................................102
4.7.2 Carrier Phase and Doppler........................................................................................104
4.8 Navigation Solution....................................................................................................... 104

CHAPTER 5

VERIFICATION OF THE SOFTWARE GPS RECEIVER BY A REAL IF GPS

SIGNAL ............................................................................................................................... 105
5.1 Software Receiver Verification Scheme...................................................................... 105
5.2 Acquisition Results Verification .................................................................................. 108
5.3 Tracking Results of the Software GPS Receiver and Navigation Solution
Verification .......................................................................................................................... 108

CHAPTER 6

SIMULATED IF GPS SIGNAL VERIFICATION RESULTS ...................................... 115

6.1 Simulator Verification Scheme.................................................................................... 115

vii
6.2 Simulated Signal Description....................................................................................... 116
6.3 Time Domain Verification............................................................................................ 117
6.4 Verification in the Correlation Domain...................................................................... 118
6.5 Tracking Result Comparison....................................................................................... 122
6.5.1 Data Bits ...................................................................................................................122
6.5.2 Carrier-to-noise Ratio ...............................................................................................125
6.5.3 Delay Lock Loop Detector .......................................................................................128
6.5.4 Doppler .....................................................................................................................131
6.5.5 Pseudoranges ............................................................................................................134
6.6 Verification in Position Domain .................................................................................. 139
6.7 Some Other Results for Various Simulation Cases ................................................... 141

CHAPTER 7

CONCLUSIONS AND FUTURE WORK ........................................................................ 151

7.1 Conclusions.................................................................................................................... 151
7.1.1 Software IF GPS Signal Simulator ...........................................................................151
7.1.2 Software GPS Receiver ............................................................................................152
7.2 Future Work.................................................................................................................. 153
7.2.1Future Work for the Software IF GPS Signal Simulator ...........................................153
7.2.2 Future Work for the Software GPS Receiver ...........................................................154

REFERENCES.................................................................................................................... 156

viii
 LIST OF TABLES

Table 2.1: Received GPS signal power levels .................................................................. 12

Table 2.2: GPS ranging errors .......................................................................................... 26

Table 3.1: Simulator parameters ....................................................................................... 75

Table 4.1: PLL discriminators .......................................................................................... 93

Table 5.1: Doppler and C/N0 of the observed satellites.................................................. 108

Table 5.2: Acquired Doppler values of the Signal Tap collected signal ........................ 108

Table 5.3: Doppler and C/N0 values obtained by the OEM4 receiver and software GPS

receiver.................................................................................................................... 112

Table 6.1: Signal simulation parameters......................................................................... 116

Table 6.2: Simulated C/N0 and ionospheric and tropospheric errors.............................. 117

Table 6.3: Correlation peak values of all acquired satellites .......................................... 121

Table 6.4 Pseudorange differences of the simulated signals and simulated ionospheric

errors ....................................................................................................................... 147

ix
 LIST OF FIGURES

Figure 2.1: Relationship between GPS signal components .............................................. 12

Figure 2.2: GPS signal description in the frequency domain ........................................... 15

Figure 2.3: Autocorrelation and cross-correlation of GPS gold codes ............................. 17

Figure 2.4: Generic GPS receiver functional block diagram............................................ 18

Figure 3.1: GPS signal propagation and processing in the front-end ............................... 31

Figure 3.2: Digitized IF GPS signal simulator structure................................................... 40

Figure 3.3: Simulated receiver trajectory modes .............................................................. 41

Figure 3.4: GPS navigation data format ........................................................................... 44

Figure 3.5: TLM and HOW word ..................................................................................... 46

Figure 3.6: C/A code generator......................................................................................... 47

Figure 3.7: Effect of Earth rotation on signal propagation ............................................... 48

Figure 3.8: Satellite and receiver position in the ECEF frame at the transmission time and

reception time............................................................................................................ 50

Figure 3.9: Description of the single-layer ionosphere model.......................................... 54

Figure 3.10: Compute TEC in a grid network .................................................................. 58

Figure 3.11: Noise characteristics in the time and frequency domains ............................ 64

Figure 3.12: Band-pass filter magnitude and phase response........................................... 65

Figure 3.13: IF GPS signal before and after filtering ....................................................... 66

Figure 3.14: IF 1-bit quantization GPS signal in the time and frequency domains.......... 67

Figure 3.15: Two-bit quantization strategy....................................................................... 68

Figure 3.16: A/D conversion gain in Guassian noise ....................................................... 70

x
Figure 3.17: IF 2-bit quantization GPS signal in the time and frequency domains.......... 71

Figure 3.18: Signal simulator functions............................................................................ 73

Figure 3.19: Program flow chart...................................................................................... 74

Figure 4.1: Structure of the software GPS receiver .......................................................... 77

Figure 4.2: Signal acquisition ........................................................................................... 78

Figure 4.3: DFT-Based acquisition................................................................................... 80

Figure 4.4: Coupled code and carrier tracking loops........................................................ 83

Figure 4.5: Basic phase lock loop ..................................................................................... 84

Figure 4.6: Delay lock loop .............................................................................................. 87

Figure 4.7: Code mismatch vs early, punctual and late components................................ 88

Figure 4.8: DLL discriminators ........................................................................................ 88

Figure 4.9: Frequency lock loop ....................................................................................... 91

Figure 4.10: Phase lock loop............................................................................................. 92

Figure 4.11: Statistics of normalized power NP for M=20............................................... 95

Figure 4.12: A successful bit synchronization histogram................................................. 96

Figure 4.13: Navigation message...................................................................................... 99

Figure 4.14: TLM word and HOW word........................................................................ 100

Figure 4.15: Pseudorange construction........................................................................... 103

Figure 5.1: Hardware front-end “GPS Signal Tap” ........................................................ 106

Figure 5.2: OEM4 receiver tracking status ..................................................................... 107

Figure 5.3 (a): Average C/N0 and Doppler for SV1 ....................................................... 109

Figure 5.3 (b): Average C/N0 and Doppler for SV2 ....................................................... 109

Figure 5.3 (c): Average C/N0 and Doppler for SV4 ....................................................... 110

xi
Figure 5.3 (d): Average C/N0 and Doppler for SV13 ..................................................... 110

Figure 5.3 (e): Average C/N0 and Doppler for SV16 ..................................................... 110

Figure 5.3 (f): Average C/N0 and Doppler for SV20...................................................... 111

Figure 5.3 (g): Average C/N0 and Doppler for SV25 ..................................................... 111

Figure 5.4: Navigation solution of the real GPS signal .................................................. 113

Figure 5.5: Navigation solution computed from the OEM4 receiver generated raw

measurements.......................................................................................................... 114

Figure 6.1: Power spectrum of the real and simulated IF GPS signal ............................ 117

Figure 6.2 (a): Acquisition results of the real signal and simulated signal for SV1 ....... 119

Figure 6.2 (b): Acquisition results of the real signal and simulated signal for SV2....... 119

Figure 6.2 (c): Acquisition results of the real signal and simulated signal for SV4 ....... 119

Figure 6.2 (d): Acquisition results of the real signal and simulated signal for SV13..... 120

Figure 6.2 (e): Acquisition results of the real signal and simulated signal for SV16 ..... 120

Figure 6.2 (f): Acquisition results of the real signal and simulated signal for SV20 ..... 120

Figure 6.2 (g): Acquisition results of the real signal and simulated signal for SV25..... 121

Figure 6.3 (a): Data bits of the real signal and simulated signal for SV1....................... 122

Figure 6.3 (b): Data bits of the real signal and simulated signal for SV2....................... 123

Figure 6.3 (c): Data bits of the real signal and simulated signal for SV4....................... 123

Figure 6.3 (d): Data bits of the real signal and simulated signal for SV13..................... 123

Figure 6.3 (e): Data bits of the real signal and simulated signal for SV16..................... 124

Figure 6.3 (f): Data bits of the real signal and simulated signal for SV20 ..................... 124

Figure 6.3 (g): Data bits of the real signal and simulated signal for SV25..................... 124

Figure 6.4 (a): C/N0 of the real signal and simulated signal for SV1 ............................. 125

xii
Figure 6.4 (b): C/N0 of the real signal and simulated signal for SV2............................. 126

Figure 6.4 (c): C/N0 of the real signal and simulated signal for SV4 ............................. 126

Figure 6.4 (d): C/N0 of the real signal and simulated signal for SV13........................... 126

Figure 6.4 (e): C/N0 of the real signal and simulated signal for SV16 ........................... 127

Figure 6.4 (f): C/N0 of the real signal and simulated signal for SV20............................ 127

Figure 6.4 (g): C/N0 of the real signal and simulated signal for SV25........................... 127

Figure 6.5 (a): DLL lock detector of the real signal and simulated signal for SV1........ 129

Figure 6.5 (b): DLL lock detector of the real signal and simulated signal for SV2 ....... 129

Figure 6.5 (c): DLL lock detector of the real signal and simulated signal for SV4........ 129

Figure 6.5 (d): DLL lock detector of the real signal and simulated signal for SV13 ..... 130

Figure 6.5 (e): DLL lock detector of the real signal and simulated signal for SV16...... 130

Figure 6.5 (f): DLL lock detector of the real signal and simulated signal for SV20 ...... 130

Figure 6.5 (g): DLL lock detector of the real signal and simulated signal for SV25 ..... 131

Figure 6.6 (a): Doppler of the real signal and simulated signal for SV1 ........................ 132

Figure 6.6 (b): Doppler of the real signal and simulated signal for SV2........................ 132

Figure 6.6 (c): Doppler of the real signal and simulated signal for SV4 ........................ 132

Figure 6.6 (d): Doppler of the real signal and simulated signal for SV13...................... 133

Figure 6.6 (e): Doppler of the real signal and simulated signal for SV16 ...................... 133

Figure 6.6 (f): Doppler of the real signal and simulated signal for SV20 ...................... 133

Figure 6.6 (g): Doppler of the real signal and simulated signal for SV25...................... 134

Figure 6.7 (a): Doppler of the real signal and simulated signal for SV1 ........................ 135

Figure 6.7 (b): Doppler of the real signal and simulated signal for SV2........................ 135

Figure 6.7 (c): Doppler of the real signal and simulated signal for SV4 ........................ 135

xiii
Figure 6.7 (d): Doppler of the real signal and simulated signal for SV13...................... 136

Figure 6.7 (e): Doppler of the real signal and simulated signal for SV16 ...................... 136

Figure 6.7 (f): Doppler of the real signal and simulated signal for SV20 ...................... 136

Figure 6.7 (g): Doppler of the real signal and simulated signal for SV25...................... 137

Figure 6.8 Pseudorange differences of the real signal and simulated signal .................. 138

Figure 6.9: Navigation solution of the real GPS signal .................................................. 139

Figure 6.10: Navigation solution of the simulated GPS signal....................................... 139

Figure 6.11: Carrier smoothed navigation solution of the simulated GPS signal........... 141

Figure 6.12: Navigation solution of the simulated GPS signal with 1-bit quantization . 143

Figure 6.13: Pseudorange and Doppler differences of the simulated GPS signal with 1-bit

quantization............................................................................................................. 143

Figure 6.14: Navigation solution of the simulated GPS signal with noise, filtering and 1-

bit quantization........................................................................................................ 144

Figure 6.15: Pseudorange and Doppler differences of the simulated GPS signal with

noise, filtering and 1-bit quantization ..................................................................... 144

Figure 6.16: Navigation solution of the simulated GPS signal with noise, filtering and 2-

bit quantization........................................................................................................ 145

Figure 6.17: Pseudorange and Doppler differences of the simulated GPS signal with

noise, filtering and 2-bit quantization ..................................................................... 145

Figure 6.18: Navigation solution of the simulated GPS signal with noise, filtering, 1-bit

quantization, and ionospheric error......................................................................... 146

Figure 6.19: Pseudorange and Doppler differences of the simulated GPS signal with

noise, filtering, 1-bit quantization, and ionospheric error....................................... 146

xiv
Figure 6.20: Navigation solution of the simulated GPS signal with tropospheric error

model applied in C3NAVG2 TM .............................................................................. 148

Figure 6.21: Navigation solution of the simulated GPS signal (receiver trajectory mode 1)

................................................................................................................................. 149

Figure 6.22: Pseudorange and Doppler differences of the simulated GPS signal (receiver

trajectory mode 1) ................................................................................................... 149

Figure 6.23: Navigation solution of the simulated GPS signal (receiver trajectory mode 2)

................................................................................................................................. 150

Figure 6.24: Pseudorange and Doppler differences of the simulated GPS signal (receiver

trajectory mode 2) ................................................................................................... 150

xv
 NOTATION

0.1 Symbols

A Signal Amplitude

A Square Root of the Semi-major Axis

Bn Bandwidth of the Filter in the Receiver to Remove the Out of Band Noise

C Pseudorandom Noise Spreading Sequence (C/A code)

c Speed of Light

D Navigation Data Bits

e Partial Pressure of the Water Vapour

Ek Eccentric Anomaly of Satellite Orbit

fd Doppler Frequency

f L1 Centre Frequency of L1

f L2 Centre Frequency of L2

f data Frequency of Navigation Data

Fe -4.442807633 × 10−10 s / m1 / 2

Gd Spectral Density of the Navigation Data

GC / A Spectral Density of the C/A Code

m(z ) Single-layer Mapping Function

xvi
N The Number of the Satellites in View

N0 Noise Density

Nd0 Dry Reflectivity at the Surface of the Earth

N w0 Wet Reflectivity at the Surface of the Earth

Pc Effective Carrier Power

Pd Dry Air Pressure

Pt Transmitted Signal Power

Pr Received Signal Power

~
Pnm Normalized Associated Legendre Functions of Degree n and Order m

R Auto Correlation Function

Rh Relative Humidity

RH 0 Average Relative Humidity

RH v Daily Change of the Relative Humidity

s Sun-fixed Longitude of the Ionospheric Pierce Point

S L1 GPS L1 Signal

SF Scale Factor

t GPS Time

toc Reference Time for Clock Correction

tT max The Epoch with the Highest Temperature

t RH max The Epoch with the Maximum Relative Humidity

∆ tR Relativistic Correction

xvii
T0 Average Temperature over a Given Day

Tc Temperature

Tp Signal Propagation Time from Satellite to Receiver Antenna

Tv Daily Change of the Relative Humidity

v
v Velocity of the Satellite
v
vu Velocity of the User

Y Encrypted P Code Spreading Sequence

z Geocentric Zenith Distance of a Satellite at the Height of GPS Receiver

z′ Geocentric Zenith Distance of the Single Layer of the Ionosphere

v
1 Line of Sight Unit Vector from User to Satellite

ωe Frequency Error

ω IF Intermediate Frequency

ω L1 L1 Carrier Frequency

ωLo Frequency of Local Oscillator

∆ω Frequency Offset

φ0 Nominal Carrier Phase

ϕ0 Initial Phase of the GPS Signal

ϕe Fixed Phase Error Offset

δ tiono Ionospheric Error

δ tr Receiver Clock Error

δ t SV Satellite Clock Error

xviii
δ ttropo Tropospheric Error

δ t MP Multipath Signal Path Delay

α Signal Power Attenuation Factor

β Geocentric Latitude of the Ionospheric Pierce Point

θ Elevation Angle

Λ nm Normalization Factor

∆s Troposphere Slant Delay

0.2 Abbreviations and Acronyms

A/D Analog to Digital

AGC Automatic Gain Control

BPSK Binary Phase Shift Keying

C/A Coarse/Acquisition

CDDIS Crustal Dynamics Data Information System Institute

C/N0 Carrier to Noise Density Ratio

C3NAVG2TM Combined Code and Carrier for Navigation with GPS and GLONASS

DLL Delay Lock Loop

ECEF Earth-Centred, Earth-Fixed

ECI Earth-Centred Inertial

FIR Finite Impulse Response

FLL Frequency Lock Loop

GPS Global Positioning System

xix
GLONASS Global Orbiting Navigation Satellite System

I In-Phase Carrier Signal

IF Intermediate Frequency

IGS International GPS Service

L1 Link 1

L2 Link 2

LNA Low Noise Amplifier

LO Local Oscillator

NCO Numerically Controlled Oscillator

PLL Phase Lock Loop

PPS Precise Positioning Service

PRN Pseudorandom Noise

P(Y) Precise

Q Quadra-Phase Carrier Signal

RMS Root Mean Square

RF Radio Frequency

SA Selective Availability

SNR Signal to Noise Ratio

SPS Standard Positioning Service

Std Dev Standard Deviation

TEC Total Electron Content

UTC Coordinated Universal Time

WAAS Wide Area Augmentation System

xx
WGS-84 World Geodetic System - 1984

xxi
 CHAPTER 1

INTRODUCTION

1.1 Background

For positioning purposes, a GPS receiver needs to fulfill several tasks to derive the raw

measurements from the GPS radio frequency (RF) signals transmitted by the satellites.

First, a well-designed antenna is necessary to receive the RF signals which are down-

converted to a low frequency band and digitized into intermediate frequency (IF) discrete

signals by an RF section, referred to as the front-end. At this point, the digitized IF signal

is ready to be processed by the receiver channels. Then, the signals are fed into different

channels allocated to each satellite. Each receiver channel conducts operations of

acquisition, tracking, and navigation message demodulation. Acquisition is used to detect

a signal and derive its initial code shift and Doppler estimates. Tracking, bit

synchronization, and sub-frame synchronization are used to keep track of the carrier and

code as well as to obtain pseudorange, carrier phase measurements, and the navigation

message. Finally, the navigation solution is computed. In a conventional receiver, the

front-end, and part of the digital signal processing (such as correlation and accumulation),

are implemented in hardware while the other part of the digital signal processing is

implemented in software.

Unlike conventional GPS receivers, a software GPS receiver is mainly realized by

software except for the front-end part, which offers various advantages. First, a software
1
approach removes the nonlinear, temperature-dependent and age-dependent components

of the hardware receiver. Second, a software receiver can provide researchers and

developers with more evaluation and testing flexibility. It is easy to prototype the latest

theoretical signal processing developments. The performance of different algorithms can

be compared without any hardware development. Third, it provides an effective

simulation environment. The processing techniques of new signals, such as L5 signals in

the GPS modernization program and Galileo signals, can be studied without any

hardware development.

All research on software GPS receivers start from the basis of the digitized IF signal.

Only a digitized IF signal with proper base band intermediate frequencies can be fed to

and processed by a software GPS receiver. Various hardware front-ends and A/D

(Analog-to-Digital) converters can be used to implement frequency down-conversion,

signal sampling, and quantization to provide a suitable IF GPS signal; however, using a

hardware front-end has several shortcomings. Firstly, a hardware front-end affords little

flexibility because it normally has its own specific and multiple, generally component-

driven, parameters that cannot be easily changed. Secondly, the use of a hardware front-

end requires a GPS simulator to generate actual radio frequency signals based on

predetermined parameters to simulate various GPS environment conditions. However, in

the typical development stage of a project, the cost of a GPS simulator is prohibitively

expensive, typically in the hundreds of thousands of dollars. Furthermore, some new GPS

signals may not be available even in the latest hardware simulators.

2
A software simulator is an efficient way to generate the digitized IF GPS signal as the

input to a software receiver. First, a software simulator allows simulation of various

signal conditions. The user can configure the GPS signal propagation and processing

chain, allowing the introduction of specialized situations to create digitized IF GPS

signals with specific properties as required. This makes it convenient to test and evaluate

various receiver acquisition and tracking algorithms and to investigate the software

receiver performance. Second, it is possible to add new signal features in a software

simulator; for example, operation with signals that are not yet available (such as L5 and

Galileo signals) can be simulated to develop and evaluate new receiver algorithms. The

flexibility and lower cost of software-based simulation provides a viable method of

generating the expected signal.

1.2 GPS Signal Simulation Review

Several software-based GPS signal simulators are currently under development, as

documented in the literature. Although the specific implementation varies between the

available simulators, they can simulate either digitized IF signals or analytical

accumulated In-phase (I) and Quadra-phase (Q) components. In some simulators, the real

GPS signals - including codes, navigation messages, and error sources - are faithfully

simulated while, in others, mathematical models are used to directly generate analytical I

and Q components with error sources without simulating the actual carrier and code

waveforms. Although the effects of errors are important, not all error sources may be

simulated in any one of the simulators due to their different research purposes.

3
Three commercial simulators were found in the literature. The MATLAB GPS signal

simulation toolbox developed by the NAVSYS Company (Brown et al., 2000) provides a

complete set of GPS signal simulations, test, and analysis tools. It can generate the

simulated RF GPS signals as a function of the signal strength, pseudorange, Doppler and

Doppler rate. The effects of interfering signal sources, antenna and front-end on the

received GPS signal are also modeled. This simulation approach is very computationally

expensive because of the simulation of RF GPS signals.

The second product found is the software-based GPS signal simulator released by the

Centre for Remote Sensing (CRS) (CRS, 2002). It is part of a suite of products developed

by CRS for the development of advanced GPS systems. It allows users to introduce

various specialized situations such as scintillation, interference and jamming, namely

narrow-band, wide-band and directional, high-dynamics; specified orbits, wave front

simulation at multiple antenna locations, plasma re-entry bodies, obscuration, multipath

effects, and ocean scattering.

The third product, available commercially through Navtech Inc., is the GPS correlator

simulator developed by Accord Software and Systems. According to Navtech, Accord’s

product is a “very powerful software package for understanding the characteristics of the

GPS satellite signal and its processing in a typical GPS receiver” (Navtech, 2002). The

simulator is composed of two parts: an input signal simulator and a signal processor. The

input signal simulator can simulate noise effects, receiver circuit parameters and the base

band converter. However, its true environment is not driven by satellite position and

4
receiver trajectory, which is a requirement for simulating ultra-tightly coupled navigation

systems.

Burns et al. (2002) presented a software GPS signal simulation method for the purpose of

investigating GPS software radio performance in combating narrow band interference.

An L1 signal was simulated including the C/A code, the navigation data, atmospheric

noise, and radio frequency interference (RFI); the performance of the antenna, filtering,

gain, down-conversion and thermal noise inherent in the analog hardware were taken into

account. Seitz and Braasch (1998) implemented a high fidelity GPS receiver simulation

that is structured to include a GPS signal generator, a propagation path that allows for the

addition of signal anomalies, and a GPS receiver model. The functions of the first two

items were similar to the simulation made by Burns et al. (2002).

Corbell (2000) developed a GPS signal simulator based on the true satellite and user

trajectories. Only the receiver errors that affect the receiver tracking loop performance

were taken into account. The errors, such as clock error, amplitude of signal and noise,

and quantization effects were modeled at the A/D output point by derivation of analytical

error expressions.

A distinctive analytical representation of a GPS receiver was also realized by Miller and

Corbell (2000) and Corbell and Miller (2001). The GPS signal simulator part does not

actually generate the carrier or code waveforms, but keeps track of their phases and

relationships to generate the analytical I and Q samples by means of mathematical

5
models. The computational burden is greatly reduced using this simulation method, but it

is less accurate at high dynamics and over long periods.

1.3 Objective

The objective of this thesis is to develop and verify a digitized IF GPS signal simulator.

This objective is part of a larger objective within the PLAN group of the Department of

Geomatics Engineering to develop a series of generic GNSS software receiver modules to

study existing and new GPS and Galileo signal performance, as well as study the effect of

various types of interference and external aiding on system performance.

The main purpose of the signal simulator developed herein is to provide a basic user-

friendly simulation tool to generate the digitized IF GPS signal for evaluation of software

receiver performance. The research is thus complementary to advanced research into

various acquisition, tracking and navigation algorithms for software GPS receivers

underway by other researchers within the PLAN group. Because of time constraints, a

basic version of the simulator, based on the L1 C/A code, is developed and verified in

this thesis.

The simulation is based on receiving time, satellites’ trajectories, user trajectory, and a

selection of GPS errors. The generated signal is saved in a data file that can be directly

read by a software receiver. The simulator endeavors to create an accurate representation

of the signals as a function of propagation channel parameters, front-end parameters, and

environmental noise. However, only those error sources that significantly affect the

6
measurements will be simulated in the first stage. To reduce the potentially large

computational burden, the error simulation is based only on the mathematical models at

the IF sample point with some key functions written in C code. Another consideration is

that the simulator must be easily reconfigured so that users can combine new components

or functions to generate the new signals and change parameters for specific signal

conditions.

Verification of the simulator’s operation is critical to the success of the project. It can be

conducted directly by feeding the simulator-generated IF GPS signal to a software GPS

receiver and observing if the correct tracking results and navigation solution are obtained.

To this end, the first step involves the development of a software GPS receiver. However,

the self-developed receiver also needs verification. Therefore, real IF GPS signals

collected by a hardware front-end at a known position is fed to the receiver to verify the

correctness of the software GPS receiver. The simulator simulates this group of real IF

GPS signals, and then feeds the results to the same software receiver. The verification is

conducted by comparing the simulated and real signals in the frequency, acquisition and

position domains and the tracking results of both signals.

1.4 Thesis Outline

The remaining parts of this thesis consist of the following chapters: Chapter 2 describes

the structure of the GPS signal, conventional front-end and receiver design, as well as

various error sources. Chapter 3 first develops an intermediate frequency GPS signal

simulation model, which includes the RF GPS signal propagation in the atmosphere and

7
processing in the front-end till the IF signal point after A/D; then the realization of the

digitized IF GPS simulator is described. Chapter 4 describes the design of a software

GPS receiver. Chapter 5 presents the software GPS receiver verification results using the

real IF GPS data collected by a hardware front-end called “Signal Tap”. Chapter 6

presents the simulator verification by comparing the simulated and real signals in the

frequency, acquisition and position domains and the tracking results of both signals.

Chapter 7 describes the conclusions and recommendations for further research.

8
 CHAPTER 2

GPS SIGNAL AND RECEIVER THEORY

2.1 GPS Signals

Currently, each GPS satellite broadcasts signals on two specific centre frequencies, L1

(1575.42 MHz) and L2 (1227.60 MHz). The transmitted signal is a modulation of the

navigation message on Pseudorandom Noise (PRN) codes and carriers. The Standard

Positioning Service (SPS) signal (C/A code) and the Precise Positioning Service (PPS)

signal (P(Y) code) are both broadcast on the L1 frequency. The C/A code is the in-phase

component, and the P(Y) code is quadrature component. Each satellite has the capability

of broadcasting either the SPS or the PPS signal on L2 but, in practice, only the PPS

signal is broadcast. According to the GPS Modernization Program, new signals such as

L5 and the M code will be implemented in the satellites of Block IIF (McDonald and

Hegarty, 2000).

2.1.1 GPS Signal Structure

Each signal consists of three components:

• Carrier: RF sinusoidal signal with frequency fL1 or fL2.

• Ranging code: The sequences, called PRN sequences or PRN codes have special

properties which allow all satellites to transmit at the same frequency without

interfering with each other. SPS uses a family of Gold codes for its PRN

9
 sequences (C/A codes), at a chipping rate of 1.023 MHz. The PPS uses a unique

segment of an extremely long (~1014 chips) PRN sequence called the P(Y) code,

which has a chipping rate of 10.23 MHz, ten times that for a C/A code. The Gold

codes are 1023 chips long, and therefore repeat every millisecond. The P(Y) code

segment repeats each week, giving it a period of 604800 seconds and a unique

sequence of 6.1871×1012 chips.

• Navigation data: this is a binary coded message consisting of data on the satellite

health status, ephemeris, clock bias parameters, and an almanac giving reduced

precision ephemeris data on all satellites in the constellation. Each GPS satellite’s

BPSK (Binary Phase Shift Keying) unit modulates all signals with a 50 Hz data

message 37500 bits long, which continuously repeats and periodically changes.

An entire data message from one satellite is received only after 12.5 minutes of

continuous reception.

These three components of a signal are derived coherently from one of the atomic

standards aboard the satellite, which is 10.23 MHz. The relationship with the chipping

rate is obvious. The relationships with L1, L2 and the navigation data rate are as follows:

f L1 = 1575.42 MHz = 154 × 10.23 MHz

f L 2 = 1227.60 MHz = 120 × 10.23 MHz (2.1)

f data = 50 Hz = 1 / 204600 × 10.23 MHz

Each code is combined with the binary navigation data using modulo-2 addition: if the

code chip and the data bit are the same (both are 0s or both are 1s), the result is 0; and if

10
they are different, the result is 1. The composite binary signal is then modulated on the

carrier by BPSK modulation.

The L1 signal leaving the antenna of the ith satellite can be represented analytically as:

S L1i = 2 Pt Di (t ) Ci (t ) cos(ω L1t ) + 2 Pp Di (t ) Pi (t ) sin(ω L1t )

(2.2)

where

S L1i is the GPS L1 signal transmitted from SVi

Pt is the transmitted C/A code signal power

Pp is the transmitted P(Y) code signal power

Di (t ) is the navigation data bits

Ci (t ) is the pseudorandom noise spreading sequence (C/A code)

ω L1 is the L1 carrier frequency

t is the GPS time

Pi (t ) is the P code or encrypted P code (Y code) spreading sequence

In order to express the BPSK signals in Equation (2.2) the binary values of the navigation

data need to be switched to ±1. A bit 0 maps into 1 and a bit 1 maps into -1. Figure 2.1

illustrates the relationships of the C/A code signal components specified in Equation

(2.2). This plot represents an idealized depiction of the signals for the purpose of

illustrating this relationship. In practice, phase transitions at these frequencies are never

as instantaneous as depicted in this figure.

11
2.1.2 GPS Signal Power Levels

GPS signals received on the Earth are extremely weak. They are well below the

background RF noise level sensed by an antenna. It is the knowledge of the signal

structure (i.e. PRN code) that allows a receiver to extract the signal buried in the

background noise and to make precise measurements. The specifications of the received

power level for users on the Earth are given in Table 2.1.

Figure 2.1: Relationship between GPS signal components

Table 2.1: Received GPS signal power levels

L1 C/A Code L1 P Code L2 P Code

Minimum received power -160.0 dBW -163.0 dBW -166.0 dBW

12
Maximum expected power -152.0 dBW -155.5 dBW -158.0 dBW

These measurements are referenced to a power level of one Watt. However, absolute

signal power is not necessary meaningful, because it does not consider the power level of

the background noise. Receiver performance is more dependent on a signal-to-noise ratio

(SNR) than the absolute signal power. SNR is a ratio of signal power to noise power as

shown in Equation (2.3). The noise power depends on the processing bandwidth of the

GPS receiver.

S
SNR = (2.3)
N 0 Bn

where

Bn is the band width of the filter in the receiver to remove the out of band noise

S is the signal power in bandwidth Bn

N is the noise power in bandwidth Bn

N0 is the noise density

In practice, the ratio of total carrier power to the noise density C/N0 in dB-Hz is the most

generic representation of signal power as it is independent of the implementation of the

receiver front-end bandwidth. The relationship of SNR and C/N0 can be represented as:

C
SNR ( dB ) = ( dB − Hz ) − Bn ( dB ) (2.4)
N0

where

C is the total received signal power

It can be seen from the above equations that the signal power and noise power can be

calculated if the noise density, receiver processing bandwidth, and C/N0 are specified.

13
2.1.3 L1 C/A Code Signal in the Frequency Domain

The GPS signal can be readily described in the frequency domain. As soon as the GPS

signal is generated, it travels through a propagation channel which contains noise and

interference. The received GPS signal is buried in the environmental noise; nothing else

but noise behaviour can be observed in the time domain for the signal before acquisition

and tracking. A description of the GPS signal in the frequency domain is definitely a

powerful tool for identifying signal properties. Since the simulator developed in this

thesis is for the L1 C/A code, the P code and L2 will not be discussed any further.

The GPS signal is generated through spectrum spreading, which is a method of taking the

data signal of bandwidth Bd that is modulated on a sinusoidal carrier to form d(t), and

then spreading its bandwidth to a much larger value Bs. Bandwidth spreading can be

accomplished by multiplying the data modulated carrier by a wide bandwidth spreading

code. The GPS signal spectrum spreading process is shown in Figure 2.2. Since the

purpose of this figure is principally to show the GPS signal power spectra density

properties in the frequency domain, no scale is given on the y axis.

The GPS navigation data bit stream with values ±1, and clock rate fd (50 Hz) is first

modulated on a carrier of power Pd to form a narrow bandwidth signal as shown in Figure

2.2 (a). This narrow bandwidth signal of bandwidth Bd is then spread in bandwidth by a

binary PRN sequence C/A code which has a clock rate fc (1.023 MHz) that greatly

exceeds the data rate. The navigation data and C/A code have the following power

spectral densities, respectively:

14
 Figure 2.2: GPS signal description in the frequency domain

1 sin(π f / f d ) 2
Gd ( f ) = [ ] (2.5)
fd π f / fd

1 sin(π f / f c ) 2
GC / A ( f ) = [ ] (2.6)
fc π f / fc

Because the timing of the data and clock transitions are synchronous, the spread spectrum

product of the navigation data and the C/A code has exactly the same spectrum as that of

the C/A code alone. Finally, the spectrum of the spreading signal is moved to the position

of fL1 as the centre frequency as shown in Figure 2.2 (b). The GPS signal next passes

through a propagation channel (assume zero channel delay for the sake of simplicity and

without loss of generality) with additive white noise, which has power spectral density

N0, to form the received signal as shown in Figure 2.2 (c). The received signal is then

15
down-converted and filtered by processing in the front-end. The resulting signal is a

filtered intermediate frequency signal as shown in Figure 2.2 (d).

2.1.4 C/A Code Auto-correlation and Cross-correlation

The definition of the autocorrelation for function f (t ) is:

+∞
R(τ ) = ∫ f (t ) f (t + τ )dt
−∞
(2.5)

The autocorrelation function of white noise is zero everywhere except at a time

difference of τ = 0 . That is, a vector of samples of a white noise process is uncorrelated

with any shifted version of itself. The PRN code sequence has a similar autocorrelation as

that of white noise. But its autocorrelation plot is periodic and the autocorrelation values

are not zeros at any time shift greater than one code chip width because the PRN

sequence is finite in length. The autocorrelation peaks occur at τ = 0 and τ = ± nTc ,

where Tc is the chip width in seconds and n is an integer value. Its second peak values

are +63 or -65 when the time shift is greater than one code chip.

Two independent, discrete, finite length white noise vectors are uncorrelated with each

other, making their circular cross-correlation values ideally zero (Maybeck, 1994). Cross-

correlations between Gold codes in the same family are not completely zero, but are

reasonably low compared to either code’s autocorrelation peak. Any two Gold codes

within the same family have the same cross-correlation values, which becomes lower

relative to a given code’s autocorrelation peak as a code gets longer (Peterson et al.,

16
1995). In the case of the C/A codes, the maximum power of any cross-correlation value is

at least –23.9 dB lower than the autocorrelation peak (only with zero Doppler offset),

(Kaplan, 1996).

The autocorrelation of PRN 1 and the cross-correlation of two GPS Gold codes (PRN 1

and PRN 2) are shown in Figure 2.3. The auto and cross-correlation plots for all other

PRNs have similar values.

Figure 2.3: Autocorrelation and cross-correlation of GPS gold codes

2.2 Conventional Hardware Front-end and GPS Receiver

There are various GPS receivers specifically designed for various range of applications. It

is very difficult to define a typical type of conventional hardware GPS receiver.

17
However, a characterization of a generic GPS receiver can still be made based on the

receiver design requirements. A generic GPS receiver functional block diagram is shown

in Figure 2.4. A GPS receiver can be broken down into various components. First, the

antenna, possibly followed by a pre-amp, receives the RF GPS signals. After the antenna,

comes the front-end that performs down-conversion, filtering, and digitization to convert

the RF signal to an intermediate frequency signal. The next section, signal processing,

includes acquisition, tracking, and navigation message demodulation to produce the

pseudorange, and carrier phase measurements, as well as ephemeris. The final component

performs the navigation processing to obtain the user position and velocity solution.

Front End

Analog IF AGC Digital IF

Preamplifier Down-Converter Filter
A/D Converter
LO

Reference Frequency
Oscillator Synthesizer

Signal Processing

Doppler Acquisition
Correlator Accumulator
Removal Tracking
Filtering
Navigation
Processing
Carrier NCO Code NCO Data
Demodulation
Measurement
Processing

Figure 2.4: Generic GPS receiver functional block diagram

18
2.2.1 Front-end

High carrier frequencies are necessary to propagate signals from space through the

atmosphere. However, these high frequencies are not easy to filter and demodulate.

Therefore, down-conversion, by mixing the incoming signal and noise with a Local

Oscillator (LO), is applied to reduce the received carrier frequency to a more manageable

value.

Usually, the receiver front-end consists of filtering and limiting, a Lower Noise Amplifier

(LNA), a frequency synthesizer, and down-conversion. Initial filtering, limiting and LNA

can be housed within the antenna to comprise an integrated antenna electronics (Van

Dierendonck, 1996). As a result, the signal after the antenna becomes an RF signal plus

narrow band noise. This signal will be mixed with the LO in the following step. Two-

stage down-conversion to IF is typical, but one stage down-conversion and even direct L-

band digital sampling have also been used (Akos, 1997). Only one stage is represented

here as an example. The processing of the GPS signal in the front-end can be depicted

analytically as Equation 2.6 (a) (b) (c) (Van Dierendonck, 1996). The errors induced by

the mixing process are included.

S IF (t ) + n IF (t ) = ( S (t ) + n(t )) ⋅ 2 cos(ω Lo t )
(2.6 a)
+ Harmonics + LO feedthrough + image noise

S IF (t ) = AC (t ) D(t ){cos((ω L1 + ω Lo + ∆ω )t + φ0 ) + cos((ω L1 − ω Lo + ∆ω )t + φ0 )} (2.6 b)

n IF (t ) = r (t ) cos(ω IF t + ϕ (t )) (2.6 c)

19
 where

A is the signal amplitude

S IF (t ) is the intermediate frequency signal

n IF (t ) is the noise after down conversion

∆ω is the frequency offset (Doppler, etc.)

φ0 is the nominal (but ambiguous) carrier phase

ω Lo is the frequency of local oscillator

ω IF is the intermediate frequency

ϕ (t) is the noise phase

When down-converting the desired signal, the frequency difference term specifies the

generated IF frequency; any other frequencies are removed from the signal through

filtering. Harmonic frequencies, generated by the mixer because it is a non-linear device,

are generally much higher than the IF frequency and are easily filtered out. Local

oscillator feed-through is a small local oscillator frequency component that “leaks

through” the demodulation process. Image noise is caused by noise energy at the image

frequencies of the mixing process, located at ω L1 − 2ω LO and ωL1 − 2ω IF . The frequency

plan of a receiver specifies all of the previously mentioned frequencies (LO, IF, carrier,

image, and harmonic frequencies) so that only the desired signal is present and the other

frequencies are rejected by filtering. The frequency plan design is key to maintaining

signal quality at the final IF stage. The performance of the filters used in the down-

conversion process is also critical to removing out-of-band interference, thereby

increasing the signal-to-noise ratio at the final IF.

20
Once the signal has been down-converted and filtered, it is ready to be sampled.

Sampling performs analog-to-digital conversion of the GPS signal. The analog signal is

typically quantized using 1 to 5-bit resolution. Five-bit resolution results in only 0.5 dB

of loss and, counter-intuitively, a greater number of bits do not yield appreciable

improvement (Van Dierendonck, 1996). An Automatic Gain Control (AGC) circuit is

used to maximize sampling resolution. The AGC scales the received signal spectrum so

that the A/D converter uses the full resolution of the A/D converter as the received signal

and noise power fluctuate. Most commercial receivers use 1-bit quantization. This

implementation does not require AGC, provided that the front-end has the required

dynamic range to accommodate gain variations and expected interference. The sampling

frequency can be selected to convert the IF frequency directly to base band or to some

residual frequency offset. The sampling frequency can also alias the signal to some

residual frequency offset by sampling the final IF frequency at less than the Nyquist rate.

This type of sampling has the effect of causing a frequency and phase reversal, due to the

high-side mixing process that occurs.

2.2.2 Signal Processing

After down-conversion and sampling, the digitized signal contains the GPS spreading and

data modulation at a reduced IF frequency. Signal acquisition must be done prior to

tracking.

GPS signal acquisition is a two-dimensional search process in which a replica code and

carrier are aligned with the digitized IF signal. The correct alignment can be identified by

21
measurement of the output power of the correlators. In other words, when both the code

phase and carrier Doppler match the incident signal, the signal is despread. Acquisition is

a coarse synchronization process giving estimates of the code phase and carrier phase.

This information is then used to initialize the tracking loops. After acquisition, control is

then handed over to the tracking loop, the function of which is to track the variation in

code phase or carrier Doppler.

The first step is signal demodulation as shown in Figure 2.4. Demodulation is the process

of removing any residual carrier and the spreading modulation via code correlation. First

the digital IF is treated to carrier removal by the replica carrier (plus carrier Doppler) to

produce in-phase (I) and quadra-phase (Q) sampled data. Note that the replica carrier

signal is being mixed with all of the GPS Space Vehicle (SV) signals and noise at the

digital IF stage. The I and Q signals at the outputs of the mixers have desired phase

relationships with respect to the desired SV. However, at this point, the code stripping

processes that collapse these signals to base band have not yet been applied. Therefore,

the I signal at the output of the in-phase mixers would consist mostly of thermal noise

multiplied by the replica digital sine wave (to match the SV carrier at IF) and the Q signal

at the output of the quadra-phase mixers would be mostly thermal noise multiplied by the

replica digital cosine wave (to match the SV carrier at IF). The desired SV signal remains

buried in noise until the I and Q signals are collapsed to base band by the code stripping

process that follows. The I and Q signals are then correlated with early, prompt, and late

replica codes (plus code Doppler). The prompt replica code phase is aligned with the

incoming SV code phase, producing maximum correlation if it is tracking the incoming

22
SV code phase. The replica carrier (including carrier Doppler) signals are synthesized by

the carrier Numerical Controlled Oscillator (NCO) and discrete sine and cosine mapping

functions and the replica codes are synthesized by the code generator, a 3-bit shift

register and the code NCO. To this point, the IF signal has been converted to base band

after carrier removal and code correlation.

Next, the base band I and Q signals are accumulated and “dumped” (latched) at a fixed

time intervals. The three in-phase components are integrated and dumped to produce

I E , I P , I L , while the three quadra-phase components similarly produce QE , QP , Q L . The

integration and dump accumulators provide filtering and re-sampling at a lower base

band rate which depend on the desired pre-detection bandwidth. The integration time is a

compromise in the design. It must be as long as possible to operate under weak signal

conditions and it must be as short as possible to operate under high dynamic stress signal

conditions.

The integrated and dumped signals are then relayed to the discriminators which provide

the noisy estimates of the errors between the input signals and locally generated replicas

due to low SNR. The classification of the discriminators is defined by the type of tracking

loop used: Delay Lock Loop (DLL), Phase Lock Loop (PLL) or Frequency Lock Loop

(FLL). A DLL discriminator outputs the code phase error of the input signal and the

replica code. A PLL discriminator produces the phase error and an FLL discriminator

provides the frequency error.

23
The loop filter performs filtering on the perceived signal demodulation errors after the

discriminator as the discriminator outputs are very noisy. The objective of the loop filter

is to reduce noise in order to produce an accurate estimate of the original signal at its

output. The filtered error estimations are then input to the carrier NCO or code NCO to

adjust the locally generated carrier and code phase to match the input GPS signals.

The spreading code and the residual carrier must be simultaneously tracked in order to

de-modulate the received signal. Code tracking is the process of using the code

autocorrelation function to maintain alignment between the code modulation in the signal

and a locally generated replica. Typically, a Delay Lock Loop is used. Code tracking

loops can be either coherent or non-coherent, which is to say that they operate with or

without carrier phase information (Van Dierendonck, 1996). Misalignments between the

received signal and the locally generated code are sensed by comparing the

autocorrelations of the parallel correlators.

To track the received carrier frequency and phase, carrier tracking should be

implemented. The carrier tracking loop can consist of either a FLL or a PLL. Frequency

tracking is more robust than phase tracking, but is not as precise as phase tracking

(Kaplan, 1996). The prompt I and Q values from the code correlation are used as the

inputs for the frequency or phase discriminator which provides a noisy frequency or

phase error measurement between the received signal and the local NCO generated

signal.

24
If lock is lost on either the code or the carrier, the receiver will lose lock on the signal and

must re-acquire it. The complete operation from signal acquisition to tracking, maintain

tracking, lose tracking, and re-acquisition is a closed loop dynamic process. The correct

code phase and carrier phase measurements can be extracted when both DLL lock and

PLL lock are achieved. Of the two signal components, the code is easier to track in

dynamic environments, due to the 293 metres equivalent range of one chip length. Carrier

frequency lock is more precise, but also more difficult to maintain than code lock, and

carrier phase lock is the most difficult to achieve due to a carrier wavelength for L1 of

only 19 cm. A phase error of 90 degrees between the received signal and the NCO

generated frequency equals to a ranging misalignment of only 4.7 cm. As a result, the

DLL is easy to lock but not precise, whereas the PLL is precise but hard to lock.

2.3 Error Sources

There are six classes of ranging errors (Parkinson, 1996) as shown in Table 2.2. A

distinction must be made between the errors that affect position accuracy and those that

affect GPS signal tracking. For example, errors that are slowly changing over time do not

adversely affect the tracking performance of a GPS receiver, but they do affect the

accuracy of the calculated position. In the thesis, not all of the listed errors are simulated

due to time restrictions. Only the errors that greatly affect software GPS receiver tracking

performance or ranging errors are simulated.

25
 Table 2.2: GPS ranging errors
Ephemeris data Errors in the transmitted location of the satellite

Satellite clock Errors in the transmitted clock

Ionosphere Errors in the corrections of pseudorange caused by ionospheric effects

Troposphere Errors in the pseudorange corrections caused by tropospheric effects

Multipath Errors caused by reflected signals entering the receiver’s antenna

Receiver Errors in the receiver’s measurement of range caused by thermal noise,

software accuracy, and interchannel biases

2.3.1 Ephemeris Errors

Uncertainties in the broadcast ephemeris produce orbital errors. These uncertainties are

due to the accuracy limitations associated with the predicted nature of the broadcast

ephemeris. Tests have shown that the orbital error is generally a few metres (Kaplan,

1996). A very effective way to handle orbital errors is to use post-processed precise

orbits. Precise orbits are derived from an extensive reference network and the accuracy

can be as high as a few centimetres. However, the ephemeris errors do not have a

significant impact on the pseudorange measurements and receiver tracking performance

relative to other errors. Thus, they are not simulated in the simulator developed herein.

2.3.2 Satellite Clock Error

Each GPS satellite carries an atomic clock with which to maintain accurate frequency and

time. These clocks have stabilities of about 1 part in 1013 over a day and they are

monitored closely and maintained by the control segment to maintain this precision

26
(Spilker, 1996b). A major source of apparent clock error is Selective Availability (SA).

Since SA has been turned off in May, 2000, the remaining satellite clock errors are due to

slowly changing signals of the clock and have no significant effect on receiver tracking.

But they do result in an offset of pseudorange. For short periods of time, the clock bias

drift can be obtained from the broadcast ephemeris. The satellite clock effect is taken into

account in the simulator.

2.3.3 Ionospheric Error and Tropospheric Error

The ionospheric error is caused by the presence of free electrons as GPS signals pass

through the upper layer of the atmosphere. The ionosphere affects the propagation speed

of the GPS signal, resulting in a time-of-arrival error at the receiver. The effects of an

ionospheric error on the signal code and carrier are different: it causes a delay on the

code, but an advance on the carrier. But the magnitudes of the delay and the advance are

equal. The effect on range may vary from 150 metres (at midday, during periods of

maximum sunspot activity, with the satellite near the horizon of the observer) to less than

five metres, (at night, during periods of minimum sunspot activity, with the satellite at the

zenith) (Luo, 2001).

The tropospheric error is caused by the refraction of the GPS signal in the troposphere.

Similar to the ionosphere in terms of effect, the troposphere also affects the propagation

speed of the GPS signal. Since the troposphere is non-dispersive at GPS frequencies, it

has the same effect on the L1 and L2 code and carrier, a net delay. The delay contains

27
two parts, the larger of which is caused by the dry atmosphere component, which is stable

and predictable. The delay resulting from water vapour is smaller, but varies greatly.

Ionospheric and tropospheric errors exhibit some spatial correlation between received

signals, depending on whether the received signals traveled through similar atmosphere

on their way to the receiver. Signals that propagate through the same elevation angles

typically share similar ionospheric and tropospheric errors. These errors can be

characterized as noisy, slowly changing biases. A model to simulate ionospheric and

tropospheric errors, developed by Luo (2001) is used in this thesis.

2.3.4 Multipath

Multipath occurs when reflected signals, in addition to the direct signal, reach the

antenna. It depends greatly on the properties of the reflector, the antenna gain pattern, and

the type of correlator used in the GPS receiver. Multipath interferes with the correlator’s

ability to precisely determine the time instant of signal reception. It affects both code and

carrier phase measurements. Code multipath is much larger than carrier phase multipath.

It can reach up to one-half of a chip length of the PRN code, assuming an environment in

which the multipath signal strength never exceeds that of the direct signal (Goldhirsh and

Vogel, 1989). By contrast, carrier phase multipath is always less than one-quarter of the

carrier wavelength (Georgiadou and Kleusberg, 1988). Typically, for static observations,

multipath is non-Gaussian in nature and shows sinusoidal oscillations with periods of a

few minutes due to the change of satellite geometry. In kinematic applications, multipath

behaves more randomly because the movement of the vehicle changes the reflecting

28
geometry in a relatively random way. Multipath errors are unique to each receiver, and

uncorrelated between signals. They are not simulated in the project presented in this

thesis due to time restrictions.

2.3.5 Receiver Noise Errors

Receiver noise errors can be divided into two categories. The first type includes those

errors related to the chosen implementation. These effects include the AGC finite

response time effects, sampling and quantization losses, losses in the code tracking

design, and hardware resolution limitations. The second type originates from the practical

limitations of component accuracy. These errors include receiver oscillator errors, filter-

induced distortions on the signal, and inherent system noise. Typically, these errors are

not completely minimized due to complexity and cost limitations. They are, rather, more

related to the hardware implementation; some errors, such as receiver clock errors caused

by the instability of the oscillator, greatly affect the receiver performance. The noise

component is incorporated in this thesis as Gaussian white noise in the implementation of

the simulator. The noise density is selected as a typical value -203 dBW/Hz. Filtering and

quantization (1-bit and 2-bit) are performed in the simulation. The simulation of the

receiver clock error is not conducted in the current version of the simulator.

29
 CHAPTER 3

IF GPS SIGNAL SIMULATOR REALIZATION

3.1 IF GPS Signal Simulation Model Derivation

An IF GPS signal simulation model is developed before any discussion of

implementation. The objective of this thesis requires the development of a software

simulator to generate the digitized IF GPS signal immediately following the front-end

and before input to the signal processing stage. The simulation model is based on this

selected simulation point and takes into account all errors introduced during the signal

propagation in the atmosphere and in the front-end.

A key consideration in developing the simulator is the simulation of various errors. The

effects of these errors have been studied in detail by the GPS community. All error

sources - specifically, errors due to the satellite clock, orbit, ionosphere, troposphere, etc.

- behave in essentially the same manner. They all produce errors on the pseudorange and

carrier phase measurements, namely, a code delay, or a carrier phase delay or advance

error. The effect on the carrier phase is also indicated by the Doppler shift. For this

reason, the simulation model is structured so as to handle any form of error by specifying

the above parameters.

30
The derivation of the simulation model is along the real GPS signal propagation path. It

starts with GPS signal generation in the satellites which is followed by signal propagation

through the ionosphere and troposphere, signal reception at the receiver antenna and,

finally, signal processing in the front-end as shown in Figure 3.1. The functional block of

the front-end consists of signal reception by the receiver antenna, down-conversion, and

IF processing which further includes filtering, sampling and digitization. All necessary

functional components are analytically simulated in the following sections.

… Satellite Clock
Ephemeris
Simulator
Ionosphere
Generate IF signal based on a
Troposphere
mathematical model
Filtering

Multipath Sampling and Quantization

Antenna
Digitized IF
Down- Sampling GPS Signal
Pre-amplifying Filtering
Conversion Digitization

Figure 3.1: GPS signal propagation and processing in the front-end

3.1.1 Transmitted GPS Signal

For the sake of clarity, the analytical derivation starts with a single ideal GPS L1 signal

as presented in Chapter 2. The L1 GPS signal transmitted by the satellite i can be

represented as (Spilker, 1996a):

S L1i = 2 Pt Di (ttr ) Ci (ttr ) cos(ω L1ttr + ϕ 0 )

(3.1)
+ 2 Pt Di (ttr ) Yi (ttr ) cos(ω L1ttr + π / 2 + ϕ 0 )

31
 where
S L1i is the GPS L1 signal transmitted from SVi

Pt is the transmitted signal power

ω L1 is the L1 carrier frequency

ttr is the signal transmit time

ϕ0 is the initial phase of the GPS signal

With minor amplitude modulation neglected (Spilker, 1996b), a satellite clock generator

produces a waveform that, in general, can be represented as:

⎡ t2 ⎤
A sin ⎢(ω0 t + ϕ 0 ) + (ϕ e + ωe t + 2π d ) + x (t )⎥ (3.2)
⎣ 2 ⎦

where

ω0 t + ϕ 0 is the desired clock phase

ϕ e + ωe t + 2π d t 2 / 2 is the causal phase error effects

ϕe is the fixed phase error offset

ωe is the frequency error

2π d t 2 /2 is the frequency drift or aging effect

x(t) is the random phase error

From Equation (3.2), it can be seen that satellite clock errors include a fixed phase bias, a

frequency error, a frequency drift, and a random phase error. To obtain high accuracy

position and time keeping, the satellite clock error is usually corrected by adjusting the

signal transmission time. The satellite clock correction term is approximated by a

polynomial δ t SV = a f 0 + a f 1 (t − toc ) + a f 2 (t − toc ) 2 + ∆t R , where a f 0 , a f 1 , and a f 2 are the

polynomial correction coefficients corresponding to the phase error, frequency error, and

frequency drift, respectively; the relativistic correction is ∆t R ; and toc is the reference

32
time for the clock correction. All of these parameters are estimated by the Control

Segment, and are contained in the ephemeris data that is transmitted to users.

With the presence of the satellite clock error, the transmitted L1 C/A code signal can be

expressed as:

S L1i = 2 Pt Di (ttr + δ t SV ) Ci (ttr + δ t SV ) cos[ω L1 (t tr + δ t SV ) + ϕ 0 ] (3.3)

where
δ t SV is the satellite clock error

Because this thesis focuses on L1 C/A code signal simulation, the P code component in

Equation (3.1) is eliminated without significant effect for purpose of the simulation

produced herein. The justification for removing this signal from the derivation is as

follows. The power of the P code signal is half of that of the C/A code signal, which is

about 20 dB below the noise floor. Also, the noise spreading properties apply to both P

code and C/A code in the same manner. Additionally, if the signal simulator filters the

incoming signal at the C/A code bandwidth, the band-pass filtering of the signal would

filter out 90% of the P code signal power. Given these facts, it is reasonable to eliminate

the P code component from the received signal for sake of simplicity of derivation. The

parameter ϕ 0 in Equation (3.2) is assumed zero for simplicity, without loss of generality.

It can be seen from the above equation that the satellite clock error causes a frequency

offset and phase offset.

33
3.1.2 Propagation through the Atmosphere to the Antenna

The signal transmitted by a satellite travels through the atmosphere before reaching the

receiver antenna. The value of the propagation delay is a major factor in the signal model.

The ionosphere and troposphere affect the GPS signal and cause errors on both code

phase and carrier phase. The GPS signal transmitted by satellite i can be represented by

the following equation with the propagation delay, ionospheric and tropospheric errors

taken into account.

S L1i = 2 Pt Di (t − T p − δ tiono − δ ttropo + δ t SV ) Ci (t − T p − δ tiono − δ ttropo + δ t SV )

cos[ω L1 (t − T p + δ tiono − δ ttropo + δ t SV ) + ϕ 0 ] (3.4)

where

t is the GPS signal receiving time

δ tiono is the ionospheric error

δ ttropo is the tropospheric error

Tp is the signal propagation time from satellite to receiver antenna

Noise, interference and multipath are also added prior to the signal’s arrival at the

antenna. The signal received by the antenna can be expressed as:

RL1i = 2 Pr Di (t − T p − δ tiono − δ ttropo + δ t SV ) Ci (t − T p − δ tiono − δ ttropo + δ t SV )

cos[ω L1 (t − T p + δ tiono − δ ttropo + δ t SV ) + ϕ 0 ]i

+ MP + noise + inteference (3.5)

where
Pr is the received signal power

34
The multipath signal is composed of the portion of the GPS signal that is reflected. The

multipath error relates to the signal path delay caused by the extra path length and the

reflection coefficient of the reflector. The effect of multipath on the direct signal can also

be shown in the code delay and carrier phase offset. In addition, it is attenuated and thus

generally has a lower power level than the direct signal. The multipath signal is

represented as:

MP = α 2 Pr Di (t − T p − δ tiono − δ ttropo + δ t SV − δ t MP ) Ci (t − T p − δ tiono − δ ttropo + δ t SV − δ t MP )

cos[ω L1 (t − T p + δ tiono − δ ttropo + δ t SV − δ t MP ) + ϕ 0 ]i (3.6)

where
α is the signal power attenuation factor

δ t MP is the multipath signal path delay

Equation (3.5) is an expression of the received GPS signal at the receiver antenna with all

observable errors included. As mentioned above, not all of these errors are simulated

herein. With the exclusion of multipath errors and interference, the resulting signal model

for the RF GPS signals that arrive at the receiver antenna is:

RL1i = 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[ω L1 (t − Td + δ tiono ) + ϕ 0 ]i

+ noise (3.7)

where
Td = T p + δ ttropo − δ t SV

Although it may not be obvious, the Doppler term is embedded in the product of

Td − δ tiono and ω L1 in Equation (3.7). The following derivation can prove this. Let us

start from the definition of the Doppler as shown in Equation (3.8) (Axelrad and Brown,

35
1996). It is equivalent to Equation (3.9) after multiplying by a time t to both sides of the

Td − δ tiono
equation and then multiplying by a factor on the right side. It can be seen
Td − δ tiono

v v v
that ( vi − vu ) ⋅ 1i is the relative velocity in the line of sight from the user to the satellite,

v v v
and thus ( vi − vu ) ⋅ 1i * t is the distance change between the user and the satellite;

Td − δ tiono is the time delay of the signal propagation between the satellite and the user,

and thus C * (Td − δ tiono ) is also the distance change between them. As a result,

v v v
( vi − vu ) ⋅ 1i * t
equals one, and thus Equation (3.10) stands. Therefore, an equivalent
C * (Td − δ tiono )

expression of Equation (3.7) is Equation (3.11) with an obvious Doppler component.

v v
vi − v u v
fd = −( ⋅ 1i ) * f L1 (3.8)
c

where
fd is the Doppler frequency
v
vi is the velocity of the satellite i
v
vu is the velocity of the user
v
1i is the line of sight unit vector from the user to the satellite

c is the speed of light

f L1 is the L1 frequency

v v v
( vi − vu ) ⋅ 1i * t
fd * t = − * f L1 * (Td − δ tiono ) (3.9)
C * (Td − δ tiono )

f d * t = − f L1 * (Td − δ tiono ) (3.10)

RL1i = 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[(ω L1 + ωd )t + ϕ 0 ]i

(3.11)
+ noise

36
 where
Td = T p + δ ttropo

ωd = 2π f d

The received signal at the antenna can be expressed as:

{ 2P }
N
RL1 = ∑ r Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cosi [ω L1 (t − Td + δ tiono ) + ϕ 0 ]
i =1

+ noise (3.12)

where
N is the the number of satellites in view

3.1.3 Down-conversion

So far, the modeled GPS signal is still at radio frequency. After the signal propagates to

the front-end and is down-converted to the IF frequency, the model will be a signal

representation at intermediate frequency. The simulation produced in this thesis is not

meant to simulate an actual front-end processing stage. The following derivation will

focus on the signal analytical expression. No matter how many stages of the down-

conversion are used in a front-end, the final output is the desired IF signal after mixing

and band-pass filtering. Assume that the local oscillator-generated signal is as shown in

Equation (3.13), and that the result of the cosine mixing is shown in Equation (3.14).

LO1 (t ) = 2 cos(ω L1t ) (3.13)

37
S IF = RL1i * LO1

= 2 Pr Di (t − Td − δ t iono ) Ci (t − Td − δ tiono )

{cos[ω L1 (t − Td + δ tiono ) − ω L1 (t + δ t r ) + ϕ 0 ]+ cos[ω L1 (t − Td + δ tiono ) + ω L1 (t + δ t r ) + ϕ 0 ]}

(3.14)

Filtering out the high frequency component, we obtain:

S IF = RL1i * LO1

= 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[ω L1 (t − Td + δ tiono ) − ω LO1 (t + δ t r ) + ϕ 0 ]

= 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[(ω L1 − ω LO1 )t − ω L1 (Td − δ tiono ) − ω LO1δ t r + ϕ 0 ]

= 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[ω IF t − ω L1 (Td − δ tiono ) − ω LO1δ t r + ϕ 0 ]

(3.15)

where
ωLO1 is the frequency of the local oscilator

ω IF is the expected intermediate frequency

δ tr is the receiver clock error

The receiver clock error, which is caused by the instability of the oscillator, is a

significant issue in receiver operation. The oscillator error can produce a frequency shift

in the received signal, and therefore a significant frequency shift also occurs in the down-

converted signal. A portion of the signal power would be filtered out when filtering is

applied because of this frequency shift. From the tracking point of view, the Doppler is

also changed. As well, the code DLL may lose track because of the misalignment of the

carrier. In light of all of these facts, the receiver clock error commands our attention

when studying receiver performance. However, due to time restrictions, this error will not

be discussed in this thesis; instead, it is assumed to be zero herein and a full treatment is

left for future work.

38
Equation (3.15) represents the IF signal for the satellite i . Finally, taking noise into

account, the composite IF signal model including all satellites in view is:

{ }
N
S IF = ∑ 2 Pr Di (t − Td − δ tiono ) Ci (t − Td − δ tiono ) cos[ω IF t − ω L1 (Td − δ tiono ) + ϕ 0 ] i
+ noise
i =1

(3.16)

3.1.4 Intermediate Frequency Processing

The IF GPS signal model has been established as shown in Equation (3.16). This model

forms the basis of the simulation discussed in this work. The composite signal then

propagates to a band-pass filter and A/D converter to be filtered and digitized and,

finally, the digitized IF GPS signal is generated. A band-pass filter is simulated, the

bandwidth of which can be selected as required. One-bit and two-bit digitization are

realized in this thesis, the details of which are described in Section 3.2.11.

3.2 Simulator Realization

The IF GPS signal simulator was developed in MATLAB, with some functions written in

C++ and called by MATLAB to accelerate speed of operation. The receiver trajectory,

satellite ephemeris, required carrier-to-noise ratio and other environment parameters are

specified to create a valid IF GPS signal. The signals of visible satellites for the receiver

as well as noise are added together to generate the combined IF GPS signal. The

functional structure of the digitized IF GPS signal simulator is shown in Figure 3.2.

39
 Noise

N satellites in view
Sampling
Signal of SV 1 Filtering
… Quantization Output
N digitized IF
GPS signal
is saved in
Amplitude Data C/A code Carrier (L1) a data file

C/N0 Broadcast Propagation time Errors:

Noise density ephemeris Doppler Satellite clock
Ionosphere
Input parameters Troposphere
Intermediate frequency Receiving time
Sampling frequency Receiver trajectory
Bandwidth Satellite position
Some other parameters

Figure 3.2: Digitized IF GPS signal simulator structure

This structure is completely based on the signal simulation model as represented in

Equation (3.16). To start a simulation, the simulation time period has to be assigned

which is referred to the signal receiving time. The receiver trajectory is simulated and the

satellite’s positions are computed based upon the simulation time and the satellite

ephemeris. Then, the signal propagation time, Doppler, and errors are calculated to

generate the required components, such as signal amplitude, navigation data, C/A code,

carrier, etc. In this thesis, satellite clock error, ionospheric and tropospheric errors are

modeled to express the signal propagation environment. The IF GPS signal for a single

satellite is the product of these components. Then, the composite IF signal is computed

by the summation of noise and the signals of all satellites in view. After filtering and

quantization, the digitized IF GPS signal samples are obtained and saved in a data file.

40
3.2.1 Satellites in View

Normally, four to seven satellites are visible for a given GPS user. Only the signals

transmitted by these satellites can reach the receiver antenna. The signal simulated in this

thesis is the combination of these signals, together with the environment noise. If the

receiver positions and satellite positions are known, the visible satellites’ PRN numbers

can be calculated.

The receiver and satellite trajectories during the simulation period are fundamental

components required to conduct the simulation. The receiver trajectory can be simulated

for different scenarios. Three trajectory modes are designed, namely mode 0, mode 1, and

mode 2. The first one is a static mode. Mode 1 is that the receiver starts at a reference

point, moving northward along a straight line with a constant velocity. Mode 2 is that the

receiver starts at a reference point, moving along a circle with a constant angular velocity.

The latter two are shown in Figure 3.3.

70 km/h 70 km/h
Start point

Start point

Figure 3.3: Simulated receiver trajectory modes

41
To compute the satellite trajectory, the satellite ephemeris must be known. In this

simulation, the broadcast ephemeris is used. It can be obtained from the website of the

Crustal Dynamics Data Information System Institute (CDDIS), one of the analysis

centres of the International GPS Service (IGS). The ephemeris file, in global compressed

RINEX navigation format, is available for any specified day. It includes the ephemeris

for all available global satellites with a 2-hour time interval.

Given the GPS time, a satellite’s position can be calculated from the satellite ephemeris

data. The satellites in view can be determined with knowledge of the simulation time,

receiver position and satellite position. Considering the fact that satellites with very low

elevation angles suffer from increased power attenuation and atmospheric errors, an

elevation angle cut-off of 10o was considered in the simulation tests. However, this value

can be easily changed to the required value by the users.

Both the satellite position and the receiver position should be in the same coordinate

system. The simulated receiver position is in the local frame, while the satellite position

computed from the ephemeris is in the Earth-centred, Earth-fixed (ECEF) frame. So, the

satellite position is converted to the local frame to be consistent with the receiver

position. Assuming that the satellite’s position in this local frame is ( x s , y s , z s ) and the

receiver’s position is (0,0,0) , the elevation angle is calculated as:

zs
θ = arctan( ) (3.17)
2 2
xs + y s

42
3.2.2 Amplitude Modeling

The signal amplitude is related to the signal power as:

A2
P= (3.18)
2

where

P is the signal power

In the simulated IF GPS signal model, the signal power is the effective carrier power of

the received signal after amplification, transmission through the cables, and filtering. The

received RF signal is transmitted from the satellite by shaped pattern antennas, and

attenuated as a result of traveling through the atmosphere as well as free space loss. In a

conventional receiver, the minimum expected carrier power at L1 for the C/A code is -

160 dBW. The effective carrier power is the combination of this power level with the

effects of antenna gain, loss from the transmission line, the filtering bandwidth, and the

amplifier gain (Spilker, 1996a).

However, receiver performance is more dependent on the signal to noise ratio rather than

on the absolute power. The carrier to noise ratio (C/N0), which is the ratio of total carrier

power to the noise power density, is frequently used in receiver design and performance

estimation; thus, it is an important value used in the simulator design presented herein.

Thus, the expected C/N0 and the noise density are defined as input parameters. The signal

power is calculated as:

C / N0 + N0

P = Pc = 10 10
(3.19)

43
 where
Pc is the effective carrier power

N0 is the noise density

3.2.3 Navigation Data Simulation

In an actual GPS system, the navigation data is a 50 Hz bi-phase coded. So, each data bit

is 20 ms long. Since the period of the C/A code is 1 ms, there are 20 C/A codes in one

data bit. Thus, all 20 C/A codes have the same phase in one data bit. Thirty data bits

make a navigation word; 10 words make a sub-frame; 5 sub-frames make a page; and 25

pages make a super frame. Figure 3.4 shows these relationships.

1 page includes 5 sub-frames

1 2 3 12.5 minutes

25 pages

4 5
1 sub-frame includes 10 words

1 2 3 4 5 6 7 8 9 10 6s

0.6 s
1 word includes 30 bits

20 ms
1 bit includes 20 C/A codes

1 C/A code 977 ns

Figure 3.4: GPS navigation data format

44
The GPS time is given by the number of seconds in one week and this value is reset every

week at the start of a week. At this time, the cyclic paging to sub-frames 1 through 5 will

restart with sub-frame 1 regardless of which sub-frame was last transmitted prior to the

start of the week. The cycling of the 25 pages will restart with page 1 of each of the sub-

frames, regardless of which page was the last to be transmitted prior to the end of the

week. All upload and page cutovers will occur on frame boundaries. Accordingly, new

data in sub-frames 4 and 5 may start to be transmitted with any of 25 pages of these sub-

frames.

The first three sub-fames of the navigation data contain all of the ephemeris parameters.

It is possible to calculate the satellite position with all of these parameters. Sub-frames 4

and 5 include almanac data which provides a truncated, reduced precision set of

ephemeris parameters. The almanac data is mainly used to aid acquisition and to give the

approximate Doppler and delay information. Each of the 25 pages of almanac data is

different. The simulation of sub-frames 4 and 5 is very complicated and not very

meaningful for this research. Therefore they are simulated only by formatting the first

two words - the TLM word and the handover (HOW) word which include the preamble,

time of the week count message, sub-frame ID and parity check bits. The formats of the

TLM and HOW words are shown in Figure 3.5. They are the same for each sub-frame.

To simulate the data of other three sub-frames, a proper ephemeris file for the user-

defined time is downloaded from the website of CDDIS, as mentioned in Section 3.2.1.

Then these parameters, combined with the proper scale factors, (Spilker, 1996b) are

45
converted to binary bits which fit the navigation data format. The first three sub-frames

are simulated with 1 and -1 sequence in the simulation.

TLM word
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
1 0 0 0 1 0 1 1
Preamble Reserved Parity

HOW word
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

TOW count message (truncated) Parity

Sub-frame ID

Figure 3.5: TLM and HOW word

3.2.4 C/A Code Modeling

The C/A code signals are generated from the product of two 1023-bit PRN sequences, G1

and G2. Both G1 and G2 are generated by 10-stage linear shift registers which are driven

by a 1.023 MHz clock. The C/A code generator is shown in Figure 3.6 (Tsui, 2000). Both

G1 and G2 shift registers have 10 stages; thus, the generated sequence length is 1023

( 1010 − 1 ). The C/A code sequence is the result of modulo-2 adding of the outputs of G1

and G2. The output of G1 is from the last bit of the shift register, while G2’s output is

generated from two bits which are referred to as code phase selection through another

modulo-2 adder.

A function for generating the C/A code was coded in MATLAB, according to Figure 3.6

and the above operation rule. It produces C/A code sequences, +1 and -1, for each

satellite with the PRN number as the input parameter.

46
 G1 Generator

Shift Register
G1
1 2 3 4 5 6 7 8 9 10

G2 C/A code
1.023MHz Reset
Clock All One Positions of these feed back
determine the satellite ID

Shift Register
1 2 3 4 5 6 7 8 9 10

G2 Generator

Figure 3.6: C/A code generator

3.2.5 Signal Propagation Time and Doppler

Intuitively, the signal propagation time is simply the distance between the transmitter and

receiver divided by the speed of light. However, the position of the receiver, which is on

the Earth, changes during signal propagation due to the Earth rotation. The effect of Earth

rotation is not a trivial issue that can be ignored by the GPS system. The following

discussion takes this into account.

For the sake of simplicity, consider the computation of the propagation time in the Earth-

Centred Inertial (ECI) frame first. Referring to Figure 3.7, there is a satellite at position

Pt at the instant a signal is transmitted and a receiver is fixed at the position Rt on the

Earth’s surface. Assume the satellite-transmitted signal propagates with speed c in a

straight line in the ECI frame. At the time of reception, the receiver has moved to the

47
position Rr in the ECI frame due to the Earth rotation. Suppose that the signal

propagation time from the satellite to the receiver at position Rr is T p ; the transmission

v
time is tt ; the reception time is t r ; and the receiver velocity is v , which is the same as

that of the Earth rotation. Equation (3.20) presents the above case from the point of view

of a non-rotating ECI frame.

v v
C Tp = rPt Rt + v Tp (3.20)

Pt

CT p

Rt
Rr

Figure 3.7: Effect of Earth rotation on signal propagation

However, the GPS application is usually in an ECEF reference frame, in which the Earth

rotates about a fixed axis with a defined rotation rate, ωe =15 "/s. This reference frame is

designated by WGS-84 (Kaplan, 1996). During the time that the signal propagates from

the satellite to the receiver, the reference coordinate frame changes because of the Earth

rotation. As the receiver is tied to the Earth, its coordinates do not change in ECEF

although its position in the ECI frame changes. The satellite position also changes

between signal transmission time and reception time.

48
The distance between the satellite position at transmission time and the receiver position

at reception time is required to compute the signal propagation time. But these two

positions straddle two time instants and, thus, two ECEF frames, ECEF frame at

transmission time (ECEFTx) and ECEF frame at reception time (ECEFRx) as shown in

Figure 3.8. Computation of signal propagation time, therefore, requires an iterative

process, using satellite coordinates from these two frames. First, let us discuss the

computation in ECEFRx referring to Figure 3.8 (b). Some assumptions have to be made:

(1) all computations and position coordinates are in ECEFRx; and (2) the satellite

position Pt f r at transmission time tt moves to Pr at reception time t r with a constant

velocity ( v x , v y , v z ) . Signal propagation time is represented by T p . Only the reception

time and receiver position at this instant are known in the simulation. The satellite

position at t r can be calculated from the time t r and the ephemeris data, and thus its

position Pt f r at tt can be expressed by Pr , ( v x , v y , v z ) and T p as shown in Equation (3.21)

which consists of four equations and four unknowns. Solving the system of equations

yields an estimate of the propagation time (and thus the transmission time) and the

approximate coordinates of the satellite at that time. However, the resulting satellite

coordinates represent the satellite position in ECEFRx. To confirm their correctness, the

computation in ECEFTx is discussed in the following paragraph.

49
 Tp

Pr SV ( x r , y r , z r , t r )
Pt f t Pt f r
vd

Tp ECEFTx Tp ECEFRx
z z
Rx ( x u , y u , z u , t r )

Rt Rt

xft yft xft yft y

x

(a) ECEF frame at transmission time (b) ECEF frame at reception time

Figure 3.8: Satellite and receiver position in the ECEF frame at the transmission
time and reception time

⎧ xt f r = x r − v x T p
⎪
⎪ yt f r = y r − v yTp
⎪
⎨ (3.21)
⎪ zt f r = z r − v zT p
⎪
⎪ ( xt f r − xu ) 2 + ( y t f r − y u ) 2 + ( z t f r − z u ) 2 = C T p
⎩

The satellite position Pt f t in ECEFTx can be calculated from the ephemeris after obtaining

the transmission time tt = t r − T p . If this position is transformed to ECEFRx, theoretically,

it should be the same as it is computed directly in this frame since Pt f r and T p should

also keep the same value. Therefore, the coordinates of the satellite position Pt f t are

50
transformed to ECEFRx with Equation (3.22). A new signal propagation time T p′ is

calculated in the same frame. Then, compare T p′ and T p ; if they have a significant

difference, a new satellite position in ECEFTx will be recomputed based on T p′ and the

ephemeris data. Then, rotate it to ECEFRx to compute T p′ again; the old value of T p′ is

saved to T p . The above steps will be iterated until T p′ matches T p within a desired range.

⎡ xt f r ⎤ ⎡ cos(ωe * T p ) sin(ωe * T p ) 0 ⎤ ⎡ xt f t ⎤
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ y t f r ⎥ = ⎢ − sin(ωe * T p ) cos(ωe * T p ) 0⎥ ⎢ y t f t ⎥ (3.22)
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢z ⎥ ⎢ 0 0 1 ⎥ ⎢ ⎥
⎣ tfr ⎦ ⎣ ⎦ ⎣ zt f t ⎦

Once the signal propagation time is obtained, the Doppler can be calculated as:

xt f r − xu yt f r − yu zt f r − zu
vd = ⋅ ( v x − v xu ) + ⋅ ( v y − v yu ) + ⋅ ( v z − v zu )
r r r

r = ( xt f r − xu ) 2 + ( y t f r − y u ) 2 + ( zt f r − z u ) 2 (3.23)

vd
f d = − f L1 ⋅
c

3.2.6 Satellite Clock Error Modeling

The satellite clock correction δ t SV can be obtained using coefficients broadcast from the

satellite after being uploaded by the GPS Control Segment. The data sets are then

transmitted in the downlink data stream to users together with ephemeris in a 2 hour

interval. Sub-frame 1 (words 8, 9 and 10) contains the data needed by users to perform

corrections of the satellite clocks. These corrections represent a second order polynomial

51
in time. Specifically, four clock correction parameters, toc , a f 2 , a f 1 , a f 0 , are provided in

the ephemeris, and the correction term is approximated by a polynomial

δ t Sv = a f 0 + a f 1 (t − toc ) + a f 2 (t − toc ) 2 + ∆t R (3.24)

where
af0 corresponding to the phase error

af1 corresponding to the frequency error

af2 corresponding to rate of change of the frequency error

∆t R is the relativistic correction

toc is the reference time for clock correction

The relativistic correction can be computed by Equation (3.25) (Spilker, 1996b). A can

be obtained from the ephemeris sub-frame 2, and E k can be calculated from the

ephemeris parameters.

∆t R = Fe A sin E k (3.25)

where

Fe -4.442807633 × 10 −10 s / m 1 / 2

A is the square root of the semimajor axis

Ek is the eccentric anomaly of satellite orbit

3.2.7 Ionospheric Error Modeling

There are many methods for ionosphere delay modeling. In this research, the objective is

to develop an ionospheric model which is valid for the simulation of the time delay along

the observation line of sight. A combined Spherical Harmonics (SPHA) and grid model is

52
used in this research, which was originally developed by Luo (2001). It is described as

follows.

The parameter of the ionosphere that produces most of the effects on GPS signals is the

Total Electron Content (TEC). Once the TEC is available, the ionospheric group delay

can be calculated as:

40.3 × TEC
δ tiono = (3.26)
cf2

where

c is the speed of light

f is the frequency of carrier

The computation of the vertical TEC involves three steps:

• Generating a global profile of the TEC distribution.

• Generating a global grid-network with additional random TEC.

• Computing the vertical TEC at any pierce point.

Among the global ionospheric models in existence, the spherical harmonics (SPHA)

model is selected to simulate the ionospheric error. The SPHA model is based on the

assumption that the ionosphere is a thin single layer. This means that all free electrons are

distributed in a spherical shell of infinitesimal thickness. Figure 3.9 shows the concept of

this model.

53
 Figure 3.9: Description of the single-layer ionosphere model

In this model, the TEC is developed into a series of spherical harmonics in a Sun-fixed

reference frame (Luo, 2001):

nmax n
~
E ( β , s ) = SF ∑ ∑ Pnm (sin β )( a nm cos ms + bnm sin ms ) (3.27)
n =0 m =0

where

E is the vertical TEC value

β is the geocentric latitude of the pierce point of the ionosphere

s is the Sun-fixed longitude of the ionospheric pierce point or sub-

ionospheric point, see Figure 3.9. It is also the difference between the

Earth-fixed longitude of the pierce point and the longitude of the Sun

nmax is the maximum degree of the spherical harmonics expansion

~
Pnm = Λ nm Pnm are the normalized associated Legendre functions of degree n and

54
 order m based on the normalization factor Λ nm  and the classical

Legendre functions Pnm

a nm bnm are the coefficients of the spherical harmonics

SF is the scale factor to control the level of TEC value in simulation

The conversion from a vertical TEC value, E , into the slant TEC value, E s , is realized by

multiplying E with a mapping function:

1 R
E s = m( z ) E = E with sinz ′ = sin z (3.28)
cos z ′ R0 + H

where

1
m( z ) = is the single-layer mapping function
cos z ′

z , z ′ are the geocentric zenith distances of a satellite at the height of the

GPS receiver and the single layer of the ionosphere, respectively

R is the radius with respect to the receiver considered

R0 is the mean radius of the Earth, approximated as 6371 km

H is the height of the single-layer of ionosphere above the Earth mean

surface, which is set to 450 km.

To use the SPHA model, the coefficients a nm bnm must be known. In this model, Global

Ionosphere Maps (GIMs) files are used. These files contain the coefficients of spherical

harmonics and other ionospheric parameters (Schaer, 1997). These GIMs can be obtained

from the Centre for Orbit Determination in Europe (CODE), one of the Analysis Centres

of the International GPS Service (IGS). Since January 1, 1996, GIMs have been routinely

55
presented as an additional product at CODE. Every day, a set of TEC coefficients is

determined. These coefficients give an approximation of the distribution of the vertical

TEC on a global scale by analyzing the so-called geometry-free linear combination of

GPS carrier phase data collected at 84 globally distributed stations.

By the time of the model was developed, the GIMs could give a spherical harmonics

expansion with a maximum degree of twelve and a maximum order of eight. To conduct

the simulation in a small area, the resolution of the TEC distribution of the ionosphere

should be greatly increased. An approach to achieve this objective is to add a high-

resolution TEC increment grid-network to the profile of the global TEC distribution. The

TEC increment can be treated as an unmodeled part of the ionospheric delay by the

SPHA model. The following describes the procedures to generate a global TEC

distribution with an enhanced resolution.

• Partition the ionosphere single-layer into a network with averagely distributed

n × m grids in the Sun-fixed frame. The vertical TEC at a certain grid

point ( β i , si ) , VTEC SPHA ( β i , si ) , can be calculated using Equation (3.28). It should

be noticed that, although the density of the grids in the network can be very high

by increasing m and n , the actual resolution of the TEC distribution has not yet

been increased, since the maximum degree and order of the model are not

changed.

• Generate the TEC increment vi , j for each grid point. vi , j is a random variable

representing the un-modeled part of the ionospheric delay. The statistical

56
 characteristics of vi , j can be described by the following distribution (Christie et

al., 1999):

f ( x ) = 1 ( 2σ ) exp( − 2 x − m σ ) (3.29)

• Build the high-resolution grid network by adding the profile of the TEC

distribution and the TEC increment. The vertical TEC at a grid point can be

computed as follows:

VTEC ( β i , si ) = VTEC SPHA ( β i , si ) + vi , j (3.30)

By introducing the TEC increment, the grid-network can represent the TEC

distribution of the ionosphere with an enhanced resolution of 360o / m × 180o / n .

The resolution of the grid network is 1.5o × 1.5o .

After building the grid network of the ionosphere with enhanced spatial resolution, the

vertical TEC value at the pierce point can be computed in two ways. First, the SPHA

model can be expanded to a higher degree and order.

In the first two steps, a high-density grid network of TEC distribution is established. The

TEC value at any point within the network can be computed using interpolation. Figure

3.10 gives the concept of the four-point grid-based algorithm to estimate the vertical TEC

at the pierce point where P is the pierce point and Gi is the surrounding grid point.

57
 Figure 3.10: Compute TEC in a grid network

The vertical TEC at P can be computed using the following equation:

4
VTEC P = ∑ wiVTECi (3.31)
i =1

where

w1 = w(t , u ) w2 = w(t ,1 − u )

w31 = w(1 − t ,1 − u ) w4 = w(1 − t , u )

w(t , u ) = (1 − t )(1 − u )
u = ( β − β1 ) /( β 2 − β1 ) t = ( s − s1 ) /( s4 − s1 )

To obtain the geomagnetic latitude, β , and longitude, s, the geographical latitude and

longitude of the pierce point and the Sun must first be calculated; then they can be

converted to the geomagnetic frame using a rotational transformation. The rotation angles

are relevant to the position of the north geomagnetic pole, which is set to ( 79.45o N,

71.71o W). The choice of the spatial weighting function w(•) is arbitrary. The algorithms

may be based on inverse distance weighted averaging, bilinear interpolation, or multi-

58
quadratic weighting functions. In the chosen model, the bilinear interpolation weighting

function is adopted according to the specifications for users of the Wide Area

Augmentation System (WAAS) (FAA, 1997; RTCA, 1998).

From the above modeling procedures, it can be seen that the characteristics of both the

spherical harmonic model and the grid model are combined.

3.2.8 Tropospheric Error Modeling

There are two major delay effects due to the troposphere. The first and larger effect is the

dry atmosphere excess delay, typically on the order of 2.3 m. The dry effect varies slowly

with local temperature and atmospheric pressure in a reasonably predictable manner. The

second effect caused by water vapour is generally smaller, 1-80 cm at zenith. Although

approximately one-tenth the size of the dry effect, the wet delay varies markedly; 10-20%

in a few hours, and is less predictable even with surface humidity measurements (Spilker,

1996c). A model based on the modified Hopfield model, which was originally developed

by Luo (2001), is described in the following.

Hopfield (1969) has developed a two-quartic zenith model of the refraction index, with

different quartics for the dry and wet atmospheric profiles. Black and Eisner (1984) have

extended this zenith model to add the elevation angle mapping function.

The tropospheric delay is caused by the larger refractive index, n( n > 1) of atmospheric

gases than that of free space, n = 1 , which causes the speed of light (group velocity) in

59
the medium to decrease below its free space value c. The increase of propagation time

caused by the troposphere can be expressed as

δ ttropo = ∫ [n( s ) − 1]ds = ∆t d + ∆t w (3.32)

TransmissionPath

where n represents both the wet and dry terms. The refractivity is defined as N=106(n-1).

The basic two-quartic model for the refractivity versus altitude h can be expressed as:

N d = N d 0 (1 − h / hd ) 4 h ≤ hd = 43 km
(3.33)
N w = N w 0 (1 − h / hw ) 4 h ≤ hw = 12 km

where

N d 0 is the dry refractivities at the surface of the Earth

N w 0 is the wet refractivities at the surface of the Earth

hd is the height of the top shell of the “dry gas”

hw is the maximum height of the wet atmosphere

The total zenith delay is then the sum of the integrated dry and wet delays along the

vertical path:

hd h 4 hw h
∆ = 10 −6 ∫ N d 0 (1 − ) dh + 10 −6 ∫ N w 0 (1 − ) 4 dh (3.34)
husr hd husr hw

where
∆ = ∆d + ∆w

⎧10 −6
⎪ N d 0 ( hd − husr ) husr ≤ hd
∆d = ⎨ 5
⎪
⎩ 0 husr > hd

⎧10 −6
⎪ N w 0 ( hw − husr ) husr ≤ hw
∆w = ⎨ 5
⎪
⎩ 0 husr > hw

60
 where

husr is the altitude of the user antenna

N d 0 and N w 0 can be expressed by the surface meteorologic data, namely,

temperature, air pressure and humidity (Spilker, 1996c)

N d 0 = 77.604( Pd / T ) Z d−1
(3.35)
N w 0 = ( e / TZ w )(64.79 + 377600 / T )

where

Tc is the temperature in Kelvin

Pd is the dry air pressure in millibars

e is the partial pressure of the water vapour in millibars

Rh is the relative humidity, which can vary from 0.0 to 1.0

Z d , Z w and e can be computed as follows:

Z d−1 = 1 + Pd [57.97 × 10 −8 (1 + 0.52 / T ) − 9.4611 × 10 −4 Tc / T 2 ] (3.36)

2 3
Z w−1 = 1 + 1650( e / T 3 )[1 − 0.01317Tc + 1.75 × 10 −4 Tc + 1.44 × 10 −6 Tc ]

17.15T − 4684
e = 6.108 ⋅ exp( )
T − 38.45

Black and Eisner (1984) built a mapping function which can describe the slant

tropospheric delay as a function of both the local elevation angle of a satellite and the

surface temperature:

M ( E , T ) = 1 / 1 − [cos E /(1 + X dw hd / Re )]2 (3.37)

For elevation angles in the range 7°<E<90°and surface temperatures in the region

− 30o C < T < 40o C , the value of Xdwhd/Re is in the range 0.00088<Xdwhd/Re<0.01.

61
Because the temperature dependence is small enough to be ignored, the above mapping

function can be simplified as follows:

M ( E , T ) ≈ M ( E ) = 1.001 / (0.001) 2 + 0.002 + sin 2 E (3.38)

Finally, the total slant delay of the troposphere can be denoted as:

∆ s = ∆ ⋅ M (E ) (3.39)

The temporal variation of the tropospheric delay is simulated in the model developed

herein. This variation is realized by adjusting the meteorological data with time. The

diurnal variations of the temperature and relative humidity are simulated with the

functions shown below:

Tv ⎛π ⎞
T (t ) = T0 + cos⎜ (t − tT max ) ⎟ (3.40)
2 ⎝ 12 ⎠

⎧ RH v π
⎪ RH 0 + 2 cos(12 (t − t RH max )) 0 ≤ RH (t ) ≤ 1
⎪
⎪
RH (t ) = ⎨ 0 RH (t ) < 0 (3.41)
⎪
⎪ 1 RH (t ) > 1
⎪⎩

where

T0 is the average temperature over a given day

Tv is the daily change of temperature

tT max is the epoch with the highest temperature

RH 0 is the average relative humidity

RH v is the daily change of the relative humidity

62
 t RH max is the epoch with the maximum relative humidity

The time used in Equations (3.40) and (3.41) is local time with units of hours. These two

functions are designed according to experimental results (Lachapelle, 1997).

3.2.9 Noise

The received noise is usually represented as white noise with constant spectral density. Its

bandwidth is considered infinite at the receiving point, but is narrowed down by several

filters during down-conversion.

In the simulation, the discrete noise vector at the output of the final band-pass filter is

directly simulated, since it is impossible to generate, via MATLAB, a continuous white

noise with an infinite bandwidth. This noise vector is band-limited white noise with the

input sampling rate. A bandwidth of 2 MHz is used in the simulation. The reasons of the

bandwidth is selected as 2 MHz are ninety percent C/A code signal power are in the 2

MHz main lobe and the bandwidth of the hardware front-end used to collect the real IF

GPS signal in this thesis is also 2 MHz. The noise power is the product of the filter

bandwidth and the noise density, another input parameter. Thus, the noise model can be

described as follows:

noise = N 0 BW n(t ) (3.42)

The power spectrum is the key quantity in the frequency domain signal analysis. Applied

to GPS signals, it can demonstrate signal power level and signal bandwidth. Applied to

63
noise, it can make clear the background noise level, interference power level, and the

interference type. Because of this, the concepts of frequency domain and power spectrum

will be introduced first, and these concepts will be used throughout this thesis to show the

signal properties in the frequency domain.

From stochastic process theory, the power spectrum of a signal and its autocorrelation

function form a Fourier transform pair. Thus, the power spectrum can be determined

from the following equation (Kaplan, 1996):

∞ ∞
S (ω ) = ∫ R(τ ) exp( − jωτ )dτ = ∫ R(τ ) cos ωτ dτ (3.43)
−∞ −∞

In the above equation, the second relationship holds as well because the autocorrelation

function is a real even function and the power spectrum is a real function. In MATLAB,

there exist several functions that can perform spectral estimation based on a finite set of

data.

(a) Noise in the time domain (b) Noise in the frequency domain

Figure 3.11: Noise characteristics in the time and frequency domains

64
The simulated noise in the time domain and its power spectrum are shown in Figure 3.11.

In the simulation, the noise density is -203 dBW/Hz, the filter bandwidth is 2 MHz, the

sampling frequency is 4.75 MHz, and the IF is 15.42 MHz. The resultant base band

frequency is 1.17 MHz.

3.2.10 Filtering

Signal filtering is completed with a MATLAB function, ‘FIR1’, which realizes a classical

windowed linear-phase Finite Impulse Response (FIR) digital filter. This filter is actually

a standard band-pass filter defined by the filter type, filter order, and pass band

bandwidth. The magnitude and phase response of the band-pass filter used in the thesis

are shown in the Figure 3.12.

Figure 3.12: Band-pass filter magnitude and phase response

65
The filter order greatly affects the performance of the filter. The higher the filter order,

the more accurate the filter, but the computation load is also higher. So, a tradeoff must

be made. An empirically-derived value of 700 is used in the simulation. It produces a

satisfactory response in the filter, as well as a tolerable computation time of about 1.6

seconds in the computer with a P4 CPU at a speed of 2.6 GHz.

The signals before filtering and after filtering in the time and frequency domains are

shown in Figure 3.13. The two plots on the left-hand side show the results before filtering

whereas the right-hand figures give the results after filtering. It is evident that the side

lobe out of pass band is removed after the signal sequence passes the filter.

Figure 3.13: IF GPS signal before and after filtering

66
3.2.11 Quantization

The fewer the number of quantization bits, the easier the digital process becomes.

However, the fewer the number of quantization bits, the greater the signal degradation

caused by quantization. For most systems it is desirable to quantize as coarsely as the

performance will permit (Krumvieda et al., 2001). One-bit and 2-bit quantization are

used in many commercial GPS receivers. A rule of thumb states that 1-bit quantization

degrades the signal-to-noise ratio by 1.96 dB and 2-bit quantization degrades the signal-

to-noise ratio by 0.55 dB (Van Dierendonck, 1996). The implementation of 1-bit and 2-

bit quantization is presented in the following paragraphs.

One-bit quantization is realized in the simulation as it is typically used in most

commercial receivers and does not need an AGC. The implementation threshold is zero

for 1-bit quantization. The resulting data value is 1 if the signal is greater than 0; and -1 if

the signal is less than 0. The 1-bit quantized signals in the time and frequency domains

are shown in the Figure 3.14.

Figure 3.14: IF 1-bit quantization GPS signal in the time and frequency domains

67
Two-bit quantization usually results in four level magnitudes data samples by three

thresholds. Any time multi-bit quantization is used, an A/D threshold control is required

(Van Dierendonck, 1996). A feedback loop is necessary to perform AGC, which is a

dynamic process. However, it is difficult to simulate this function without a dynamic

simulation tool such as ‘Simulink’. Therefore, another method, developed by Amoroso

(1983), is used to conduct 2-bit quantization for the software implementation. In this

method, the high magnitude samples are given much more weight in the data detection

process than the small magnitude samples to achieve good performance in terms of

Gaussian noise. The general quantization strategy is illustrated in Figure 3.15. The sign

threshold is indicated on the input axis as T0 , and the positive and negative magnitude

thresholds are indicated as T0 + ∆ and T0 − ∆ , respectively. Samples falling below the

magnitude thresholds are given unity weight, while samples exceeding the magnitude

thresholds are given weight R .

Output

1
Input
T0 − ∆ T0 T0 + ∆
−1

−R

Figure 3.15: Two-bit quantization strategy

68
In the next step, one computational criterion is made for setting the ∆ and R values. The

received GPS signal is present amidst a background of Gaussian noise. The signal-to-

noise ratio of the A/D conversion input signal is extremely low. A high A/D conversion

gain is expected because the channel bit voltage will be the sum of a large number of

quantized chip voltages and will therefore tend to be Gaussian distributed with high

processing gain according to Amoroso (1983). Hence, the criterion is to set ∆ and R to

maximize the A/D conversion gain.

Assuming that the input signal is ± Vi and the Gaussian noise with zero mean has a

standard deviation of σ i , the A/D conversion gain, G , can be presented by the following

equation:

2
⎡ 2 ⎛∆⎞ ⎤
⎢ + 2Φ ⎜⎜ ⎟⎟( R − 1)⎥
lim G
⎛ ∆⎞
⎜⎜ R, ⎟⎟ = ⎣ π ⎝σi ⎠ ⎦
⎝ σi ⎠ ⎡ ⎛ ∆ ⎞⎤ ⎛∆⎞
Vi
→0
σi R 2 ⎢1 − α ⎜⎜ ⎟⎟⎥ + α ⎜⎜ ⎟⎟ (3.44)
⎣ ⎝ σ i ⎠⎦ ⎝σi ⎠

⎛ ∆⎞
= G0 ⎜⎜ R, ⎟⎟
⎝ σi ⎠

where

u 1 − y2
α (u ) = ∫ Φ ( y )dy, Φ( y ) = e 2
−u
2π

The value of Vi σ i is typically very small in a system with high processing gain. G0 is

independent of the input signal level Vi as long as the level is small. Therefore, the A/D

converter acts almost as a linear device which degrades the SNR by a ratio depending on

∆ σ i and R . The conversion gain is plotted in Figure 3.16 with ∆ σ i reinterpreted as the

69
percentage of input samples falling at a high magnitude, reflecting the adaptive threshold

strategy. This percentage is essentially independent of the input signal level as long as

⎡ ∆ ⎤
that level is low. It is defined as 100 × ⎢1 − ∫ σ∆i e − y 2 dy ⎥ .
2

−
⎢⎣ σi ⎥⎦

R=3.5
R=3
R=4
A/D Conversion Gain (dB)

R=3
R=4

R=2

R=1000

Percent of Chips at High Magnitude

Figure 3.16: A/D conversion gain in Guassian noise

Figure 3.16 shows that the joint optimum condition is substantially reached at

R = 3.5 and 35 percent of chips having high magnitude. The nearly optimal state occurs

when R is 3 or 4.

For expression simplicity, the 2-bit quantization in this thesis takes the values R = 3 and

30 percent of chips at high magnitude to determine the threshold. The sign threshold T0 is

set to 0 since the Gaussian noise is zero mean. The MATLAB quantization routine

70
iteratively determines the threshold values to meet 30 percent high magnitude sample

distribution goals. The resultant values for the digitized signal are +3 when the signal is

greater than the positive threshold; +1 when the signal is between zero and the positive

threshold; -1 when the signal is between the negative threshold and zero; and -3 when the

signal is less than the negative threshold. The resultant digitized IF GPS signal is shown

in Figure 3.17.

Figure 3.17: IF 2-bit quantization GPS signal in the time and frequency domains

3.2.12 Program Flow Chart

All of the above functions and models are realized by MATLAB functions. A block

diagram of the code of the signal simulator, structured with respect to the hierarchy of

functions, is shown in Figure 3.18. The order in which the functions are called is

predominantly from left to right. A more detailed program flow chart is given in Figure

3.19.

In summary, the simulator has the capabilities of simulating the L1 C/A code and data

modulation, ionospheric and tropospheric errors, noise, filtering and quantization. Any of

71
these functions can be turned on or off simply by a switch variable. The switch variables

as well as most of the system parameters are considered in the simulation configuration

function ‘parameters’. To simulate a specific scenario, a user simply reconfigures the

‘parameters’ function as required before simulation. The parameters included in the

‘parameters’ function are described in Table 3.1.

72
 Parameters

Simulator RxTrajectory

SatTrajectory (C)

SatValid

GenData (C)

GenComSig Signal_prn CompError Tp_doppler(C)

Quantizer FiltSig Create_code_samples IonoError(C)

GetData TropoError(C)

Figure 3.18: Signal simulator functions

73
 Simulator

Start GenComSig
Signal for satellites Signal_prn
Load parameters in view:
RxTrajectory Compute errors
Signal for SV1
+ .
C code Rx trajectory . Carrier
Compute satellite .
position according the Satellites trajectory + Signal for SV N create_code_samples
Code
broadcast ephemeris
parameters
Data GetData
Noise
SatValid Satellites in view

C code Band-pass filter Amplitude

Generate navigation data
Convert broadcast
for each satellite
ephemeris data to Amp*Data*Code*Carrier
navigation data format End
Open signal file End

N CompError
t< runtime? s
FiltSig Satellite clock C code
Y End
Compute Tp and
Set time for one loop Doppler
Tp Doppler
Interpolation
Generate IF signal
Ionosphere IonoError (C)
Quantization
Quantizer Troposphere TropoError (C)
Save Quantized IF signal
End
t + time for one loop

Figure 3.19: Program flow chart

74
 Table 3.1: Simulator parameters
Parameter Name Description
GPS System Constants:
Frequency L1 GPS L1 frequency 1575.42 MHz
ChipRate C/A code chipping rate 1.023 MHz
LightSpeed Speed of light 299792458 m/s
Simulation Control Parameters:
AddAmp Simulate signal amplitude
AddData Simulate navigation data
AddCode Simulate C/A code
AddCarrier Simulate Carrier
AddNoise Simulate noise
ApplyDopplerTp Simulate Doppler and propagation time
ApplyIono Simulate ionosphere error
ApplyTropo Simulate troposphere error
ApplyBandFilt Conduct signal filtering
QuantSig Conduct quantization
Simulation System Parameters:
StartTime Simulation start time in GPS time (s)
EndTime Simulation end time in GPS time (s)
Loop_time Simulation time of each loop
RxMode Receiver trajectory mode (0, 1, 2)
TrajTD Receiver trajectory time interval
ElevCutoff The cutoff elevation angle
NoiseDensity Expected front-end noise density
C/N0 Expected C/N0
FE_BW Band-pass filter bandwidth
FE_filter_order Filter order
QuantBit Quantization bit 1 or 2

75
 CHAPTER 4

SOFTWARE GPS RECEIVER

4.1 Introduction

A software GPS receiver becomes the advisable option if the validity of the signal

simulator needs to be verified. Partially to this end, a post-mission multiple channel GPS

receiver was developed with the functional diagram shown in Figure 4.1. Just like a usual

GPS receiver, this software receiver is composed of signal acquisition, signal tracking,

sub-frame synchronization, and navigation processing components. To enable successful

operation, a few assisting modules are also developed; for example, a measurement

derivation module to obtain pseudorange, carrier phase and Doppler from each channel;

an inner clock generation module to maintain an inner GPS timer; and a receiver

monitoring module to manage receiver channels.

The simulated signal samples are saved in an input data file. After reading these samples

into memory, the acquisition stage is engaged first to provide the initial code offset and

Doppler estimates. With these initial values, a tracking loop is able to keep track of the

simulated signals and to provide pseudorange and carrier phase estimates. Sub-frame

synchronization is used to demodulate the navigation message, to obtain satellite

76
ephemeris, and to remove the transmit time ambiguity. Finally, the user position is

computed and the inner clock bias is removed to improve the quality of the pseudorange

measurements in the navigation solution stage.

Acquisition Software GPS Receiver

IF GPS Signal Tracking Loop

Simulator Bit
Synchronization
Digitized IF GPS signal
inputs to the software Sub-frame
receiver to be verified Synchronization

Measurement Navigation
Derivation Solution

Inner Clock
Generation

Receiver
Monitoring

Figure 4.1: Structure of the software GPS receiver

4.2 Signal Acquisition

Acquisition is a coarse synchronization process giving rough estimates of the PRN code

offset and the carrier Doppler which initialize tracking loops for signal tracking and

navigation data demodulation (Lin and Tsui, 2000). As illustrated in Figure 4.2, GPS

signal acquisition is basically a two-dimensional search process in which a replica code

and a replica carrier are aligned with the received signal. The correct alignment is

identified by measurement of the output power of the correlators. In other words, when

both the code and the carrier Doppler match the incident signal or simulated signal, the

77
signal can be despread and the data can be demodulated. The results of the two-

dimensional search consist of an estimate of the code offset to within half a chip and a

Doppler estimate to within the lock range of the tracking loops.

IF GPS signal In-phase

Correlation I_corr2
Power
values

Quadra-phase
Correlation Q_corr2

90o shift

Local carrier Replica code

Frequency bins

Figure 4.2: Signal acquisition

The two-dimensional search space covering the full range of the uncertainty of code and

Doppler needs to be defined before acquisition can be performed. It is difficult to

predetermine the code offset because this offset is a function of the starting point and

changes at the chipping rate with time. Thus, the code search space typically includes all

possible code offset values. The resolution of the code search needs to be smaller than

half a chip; usually, the sampling frequency is used to specify this resolution. By

comparison, the change of Doppler is a function of user dynamics and the stability of the

receiver oscillator which is normally a low frequency quantity and easily estimated.

Therefore, the frequency searching space can be reduced if an initial estimate of Doppler

is known in advance. Without such an initial estimate, the Doppler searching space is

usually from -5 kHz to 5 kHz. The frequency resolution is determined by the coherent

78
integration time (or dwell time). The relationship is roughly ∆f = (2 3)T (Raquet, 2001)

where ∆f is the frequency bin width in Hz and T is the predetection integration time in

seconds. There is a tradeoff between the predetection integration time and acquisition

speed. Longer integration can provide better frequency resolution and higher sensitivity,

but it needs to search a greater number of bins and requires more time.

Various acquisition methods with search and detect strategies have been proposed in the

literature (Krumvieda et al., 2001; Kaplan, 1996). In a conventional receiver, acquisition

is performed using a carrier and a C/A-code replica. It resides physically on each bin for

a predetermined dwell time and, thus, the acquisition time is the product of the dwell time

and the number of search bins. Consequently, the acquisition time is very long. In a post-

mission software GPS receiver, the samples of the incident signal are first read into a

large buffer, making block data processing possible. A popular block signal acquisition

technique is the Discrete Fourier Transform (DFT) based technique which is able to

search all possible code offsets in one DFT-based computation, dramatically reducing the

computational burden (Tsui, 2000).

4.2.1 DFT-based Acquisition

Figure 4.3 demonstrates the concept of the DFT-based acquisition method. The local data

is a combination of the carrier and the C/A-code. The circular convolution gives the

acquisition results of all possible code offsets at a specific carrier frequency.

79
 Input data

Local carrier at f1 Circular Convolution Output 1

Local C/A code

Circular Convolution Output 2

Local carrier at f2

Local C/A code

Circular Convolution Output n

Local carrier at fn

Local C/A code

Figure 4.3: DFT-Based acquisition

Let the input signal be expressed as

s = AC cos(2πft ) , (4.1)

and the local carrier at f i be expressed as

li = exp( j 2πf i t ) (4.2)

Thus, their product is

AC AC
pi = exp( j 2π ( f + f i )t ) + exp( j 2π ( f i − f )t )
2 2
AC AC
= exp( j 2π ( f + f i )t ) + exp( j 2π∆ft ) (4.3)
2 2
= piH + piL

The correlation result zi between pi and the local C/A-code, CCA , gives the final

acquisition result. Since pi contains both high and low frequency components, zi can be

expressed as the summation of the two parts

80
 zi = ziH + ziL (4.4)

where

ziH is the correlation between piH and CCA , and

ziL is the correlation between piL and CCA

Written in component form, ziL is equal to

N −1
ziL = ∑ AC (n )exp( j 2π∆fTn )CCA (n + m )
n =0

⎧ ⎛ 1 − exp( j 2π∆fTN ) ⎞
⎪A⋅ ⎜ ⎟ m=0
= ⎨ ⎜⎝ 1 − exp( j 2π∆fT ) ⎟⎠ (4.5)
⎪⎩ sufficiently small m≠0
⎧ ⎛ Sa (π∆fTN ) ⎞
⎪ A ⋅ N ⋅ exp( jπ∆fT (N − 1))⎜⎜ ⎟⎟ m = 0
=⎨ ⎝ Sa (π∆fT ) ⎠
⎪⎩ sufficiently small m≠0

Obviously, ziL is significantly large only when the incident code is synchronized with the

local code and when the local carrier frequency is close to the incident carrier. ziH is

always small because it is the integration of a higher frequency component.

Such correlation can be computed by circular convolution because of the following

property (Van Vee and Coenen, 1991): the correlation between two periodic sequences

x (n ) and h (n ) ,

N
z (n ) = ∑ x (m )h (n + m ) (4.6)
m =0

equals

( )
z (n ) = IFFT DFT ( x (n )) ⋅ DFT (h (n )) , (4.7)

81
where (⋅) denotes a complex conjugate. This method can give the correlation value at all

possible code offsets in one computation. Fast acquisition speed can be achieved if the

FFT technique is applied, a routine that greatly reduces the computational burden. For

this reason, a DFT-based acquisition method is used in this software GPS receiver.

4.3 Signal Tracking

Acquisition produces a coarse estimate of the carrier Doppler and the code offset of the

incoming signals. Control is then handed over to the tracking loops, the function of which

is to track variations in the carrier Doppler and code offset due to line-of-sight dynamics

between satellites and the receiver. Together with bit and sub-frame synchronization,

precise pseudorange and carrier phase measurements can be derived. Another important

function of the tracking loop is to demodulate the navigation data to obtain the ephemeris

from which satellites positions can be calculated.

To track an incoming GPS signal, both the carrier phase and code offset need to be

matched by the locally generated carrier and code. Thus, the lock status of both the

carrier lock loop (FLL or PLL) and delay lock loop are required for signal tracking; they

must be coupled together as shown in Figure 4.4 (Tsui, 2000). Once one of them loses

lock, the other one will subsequently lose lock as well. Before discussing the delay lock

loop and the carrier lock loop implemented in the developed software GPS receiver, the

basic concept of a phase lock loop is first described in the following section.

82
 Delay locked loop (DLL)
C/A Code Integration &
Dump
Early I_E Q_E
Code
Loop filter Discriminator
generator
Late I_L Q_L
Integration &
Carrier Dump

IF GPS signal

Phase locked loop (PLL)

Prompt Integration &
Dump
I_P
Carrier
Loop filter arctan
NCO
90o Q_P
CW Integration &
Dump

Figure 4.4: Coupled code and carrier tracking loops

4.3.1 Basic Phase Lock Loop

In this section, the basic concept of the phase lock loop is described, which includes the

transfer function, the error transfer function, and the noise bandwidth.

The main purpose of a PLL is to adjust the frequency or phase of a local oscillator to

match the frequency or phase of an input signal, referred to as the reference signal. The

basic phase lock loop is shown in Figure 4.5.

83
 θi(t) ε (t) VC(t) VO(t)
Σ k0 LP Filter
+ _
θf(t)

VCO

(a) Time domain

θi(s) ε (s) VC(s) VO(s)

Σ k0 LP Filter
+ _
θf(s)

VCO

(b) S domain

Figure 4.5: Basic phase lock loop

Figure 4.5(a) shows the time domain configuration and Figure 4.5(b) shows the S domain

configuration, which are related by the Laplace transform. The input signal is θ i (t ) and

the output from the voltage controlled oscillator (VCO) is θ f (t ) . The phase comparator

measures the phase difference between these two signals. The amplifier, k 0 , represents

the gain of the phase comparator and the low-pass filter limits the noise in the loop. The

input voltage V0 to the VCO controls its output frequency, which can be expressed as

ω2 (t ) = ω0 + k1V0 (t ) (4.8)

where ω0 is the centre angular frequency of the VCO and k1 is the gain of the VCO. The

phase angle of the VCO can be obtained by integrating Equation (4.8) as

∫ ω (t )dt = ω t + ∫ k V (t )dt
t t
2 0 1 0 (4.9)
0 0

84
where

θ f (t ) = ∫ k1V0 (t )dt
t
(4.10)
0

The Laplace transform of θ f (t ) is

k1
θ f (s ) = V0 (s ) (4.11)
s

From Figure 4.5b, the following equations can be written.

VC (s ) = k 0ε (s ) = k 0 [θ i (s ) − θ f (s )] (4.12)

VO (s ) = VC (s )F (s ) (4.13)

From these three equations, one can obtain

VC (s ) VO (s ) sθ f (s )
ε (s ) = θ i (s ) − θ f (s ) = = = (4.14)
k0 k 0 F (s ) k 0 k1 F (s )

or

⎛ s ⎞
θ i (s ) = θ f (s )⎜⎜1 + ⎟ (4.15)
⎝ k 0 k1 F (s ) ⎟⎠

where ε (s ) is the error function

The transfer function H (s ) of the loop is defined as

θ f (s ) k 0 k1 F (s )
H (s ) = = . (4.16)
θ i (s ) s + k 0 k1 F (s )

The error transfer function is defined as

ε (s ) θ i (s ) − θ f (s ) s
H e (s ) = = = 1 − H (s ) = (4.17)
θ i (s ) θ i (s ) s + k 0 k1 F (s )

85
The equivalent noise bandwidth is defined as

∞
Bn = ∫ H ( jω ) df
2
(4.18)
0

where ω is the angular frequency and is related to the frequency f by

ω = 2πf (4.19)

It is evident from the above discussion that the loop filter, F (s ) , is the key component of

a PLL. It determines the loop order, natural frequency, damping coefficient, and noise

bandwidth, and thus greatly affects the overall performance. In the software receiver

discussed below, a three order phase lock loop and a two order DLL are implemented by

careful loop filter selection and parameter adjustment.

4.3.2 Delay Lock Loop

Inside a GPS receiver, a DLL is used to track the C/A code of incident signals. As in a

regular phase lock loop, it consists of a code phase discriminator, a loop filter, and the

C/A code NCO. As illustrated in Figure 4.6, the DLL realized in the software receiver is

a second order phase lock loop which tolerates constant acceleration.

86
 IE/QE I E2 + Q 2
− I L2 + Q 2
Integration E L
Correlation
& Dump I E2 + Q 2
+ I L2 + Q 2
IL/QL E L

DLL discriminator

a2ω0 (ω0)2

E/P/L
code 1
S

Code NCO

Nominal Doppler aiding

chipping rate from carrier loop

Figure 4.6: Delay lock loop

As mentioned before, the carrier and code need to be matched simultaneously by the

locally generated signal in order to keep track of the incoming signal, thus after Doppler

removal, the input of a DLL is nearly a pure C/A code (together with the navigation bits).

The correlation process mixes this input with the local early (E), punctual (P), and late

(L) code replicas in both the In-phase and Quadrature-phase arms to get the IE, QE, IP, QP,

IL, and QL. As shown in Figure 4.7, the correlation power or the correlation envelope of

the early component and the late component contains the code mismatch information ∆τ ;

thus, the IE, QE, IL, and QL are usually used in the DLL discriminator to obtain the

estimate of ∆τ . Several DLL discriminators have been proposed, and their input-output

relationships are illustrated in Figure 4.8. A normalized early minus late envelope DLL

discriminator is used in the software receiver because its tracking range is (-1.5 chips, 1.5

chips), which is much wider than that of other types of discriminators; its input-output

87
relationship is linear among (-0.5 chips, 0.5 chips), so no extra approximation will be

introduced in the estimation of ∆τ .

RP(τ)
D(τ) = RE(τ) - RL(τ)
RL(τ)
RE(τ)

∆τ

-d -d/2 d/2 d -d -d/2 d/2 d

Figure 4.7: Code mismatch vs early, punctual and late components

Figure 4.8: DLL discriminators

External aiding from the carrier loop is normally applied to improve DLL accuracy as

shown in Figure 4.6. In this case, a DLL needs only to correct some errors, such as initial

88
tracking errors, the rate of change of the ionosphere or differences in code, and carrier

multipath. These residual variations are ordinarily quite small and change very slowly

with time; thus, the DLL loop bandwidth can be significantly reduced – approaching a

magnitude on the order of 0.05 ~ 1 Hz, depending on the application. With external

aiding from the carrier loop, the pre-detection time can be extended dramatically based

on the Doppler aiding accuracy. Based upon the above analysis, a PLL aided DLL is

realized and a 2 Hz bandwidth and a 10 ms coherent integration time are selected in the

software receiver. It can achieve the satisfactory DLL tracking accuracy for the normal

user dynamic situation with a speed of tens of kilometers per hour.

A first-order loop filter is used here to construct a second-order DLL loop which can

track constant acceleration in an unbiased manner. The transfer function of the loop filter

is

ω02
F (s ) = a2ω0 +
s (4.20)

The DLL transfer function can be derived as follows after substituting F (s ) into

Equation (4.16) and assuming k 0 = 1 and k1 = 1 ,

a 2ω0 s + ω02
H (s ) = (4.21)
s 2 + a 2ω0 s + ω02

Obviously, the natural frequency of the loop is ω0, and the damping factor is 0.5a2. The

DLL noise bandwidth, Bn, the damping factor 0.5a2, and the natural frequency are related

via the following equation for a second-order loop

ω0 (1 + a 22 )
Bn = (4.22)
4a 2

89
To make the DLL a critically damped loop, a 2 needs to be equal to 1.414. Therefore,

Bn = 0.53ω0 can be derived and all of the parameters related to designing a DLL are thus

determined. The output of the loop filter is a variation of the code chipping rate. Together

with the Doppler aiding from the carrier loop and the nominal chip rate, it can drive the

code NCO properly to keep track of the code of the incoming signal and to provide an

accurate pseudorange measurement.

4.3.3 Carrier Tracking Loop

There are two types of carrier tracking loops, FLL and PLL. An FLL keeps track of the

frequency of the incoming signal by measuring the frequency mismatch between the local

carrier and the incoming carrier, while a PLL keeps track of the phase of the incoming

signal by measuring the phase mismatch between the local carrier and the incoming

carrier. In general, a PLL is more accurate than an FLL, but it is more sensitive to

dynamic stress. As a result, an FLL is normally used in high dynamic situations or in

transition stages from signal acquisition to phase tracking where it is difficult to acquire

the phase of the carrier because of large frequency uncertainties. An FLL is also less

sensitive to data bit transition than a PLL (Van Dierendonck 1996). In the software

receiver, a FLL is designed to keep working until fine phase tracking can be done and bit

synchronization is achieved. After that, a PLL will take over the carrier tracking in order

to provide accurate carrier phase and Doppler measurements.

The second-order FLL used in the software GPS receiver is illustrated by Figure 4.9.

90
 IP sign ( dot ) cross
Doppler Integration
Removal & Dump QP ( t 2 − t1 )

FLL discriminator

cosφk sinφk
a2ω0 (ω0)2

1
S
Carrier
NCO

Figure 4.9: Frequency lock loop

The key characteristic of an FLL is its discriminator which detects frequency mismatch.

In GPS receivers, frequency mismatch is usually estimated by investigating the phase

rotation between two consequential (Ip, Qp) pairs. There are several FLL discriminators

(Kaplan, 1996), such as the cross-product discriminator,

δf i = I i −1Qi − I i Qi −1 (4.23)

the decision-directed cross-product discriminator,

δf i = (I i −1Qi − I i Qi −1 )sign(I i −1 I i + Qi Qi −1 ) (4.24)

and the differential arctangent discriminator,

⎛ Qi ⎞ ⎛Q ⎞
δf i = arctan ⎜⎜ ⎟⎟ − arctan ⎜⎜ i −1 ⎟⎟ (4.25)
⎝ Ii ⎠ ⎝ I i −1 ⎠

In the software GPS receiver, the decision-directed cross-product discriminator is used

because it is insensitive to data bit transitions.

91
A PLL is much more accurate in carrier tracking than an FLL because it tracks the phase

instead of the frequency; thus, receiver channels usually transform from an FLL state to a

PLL state after the carrier mismatch converges into the lock range of the PLL used. The

PLL realized in the software GPS receiver is a third-order PLL which is illustrated in

Figure 4.10.

Doppler Integration IP ATAN ( Q PS / I PS )

Removal & Dump QP

PLL discriminator

cosφk sinφk
b3ω0 a3(ω0)2 (ω0) 3

1
S

1
S

Carrier
NCO

Figure 4.10: Phase lock loop

As mentioned before, the fundamental difference between a PLL and an FLL is the

nature of the discriminator. The discriminator of an FLL detects frequency mismatch

while the discriminator of a PLL detects phase mismatch. As listed in Table 4.1 (Raquet,

2001), there exist quite a few PLL discriminators which investigate the phase angle of the

punctual in-phase component and quadrature-phase component pair ( I P , QP ).

92
 Table 4.1: PLL discriminators
Discriminator Costas Output phase Notes
Algorithm Loop error
Sign(I)·Q Y sinφ Near optimal at high SNR. Slope
proportional to signal amplitude.
I·Q Y sin2φ Near optimal at low SNR. Slope
proportional to signal amplitude squared.
Q/I Y tanφ Suboptimal, but good at high and low
SNR. Slope is not signal amplitude
dependent.
atan2(Q,I) N φ Optimal (maximum likelihood estimator)
at high and low SNR. Slope is not signal
amplitude dependent.
atan(Q/I) Y φ Optimal (maximum likelihood estimator)
at high and low SNR. Slope is not signal
amplitude dependent.

In the software receiver, the atan (Q I ) discriminator is used because of its high

performance. The loop filter transforms phase mismatch into a carrier Doppler change

which drives the carrier NCO to reduce the phase error. Without external aiding, such as

aiding from an Inertial Measurement Unit (IMU), a PLL needs to keep track of the total

dynamic stress, such as user dynamics, inner oscillator error, atmospheric effects, and

multipath effects. This makes the carrier loop a weak loop which may lose lock easily. As

a result, a third-order PLL is realized here because it can tolerate higher dynamics. A

high order DLL is not necessary because a DLL is normally assisted by the carrier loop

which can remove user dynamics and variation of the receiver oscillator, the major

93
sources of loop dynamic stress. A DLL simply needs to track the difference between the

code and carrier loops which is quite small.

4.4 DLL Lock Detectors and C/N0 Estimation

Knowledge of code lock is obviously the same as the knowledge of received signal power

(Van Dierendonck 1996). Code lock detection is very similar to estimating C/N0. A good

C/N0 would be required only if the receiver is operating on or near the correlation peak,

say in a situation where code lock is achieved.

Consider measurements of total power in 1/T (wide band power) and 1/MT (narrow-band

power) noise bandwidths of the following form:

⎛M ⎞
WBPk = ⎜ ∑ ( I i2 + Qi2 ) ⎟ (4.26)
⎝ i =1 ⎠k

2 2
⎛M ⎞ ⎛M ⎞
NBPk = ⎜ ∑ I i ⎟ + ⎜ ∑ Qi ⎟ (4.27)
⎝ i =1 ⎠ k ⎝ i =1 ⎠ k

The exact relationship between these measurements and signal-plus-noise power is not

known. However, a normalized power defined as Equation 4.28 gives statistics that

provide a monatomic function of C/N0 which is shown in Figure 4.11 (Van Dierendonck,

1996).

NBPk
NPk = (4.28)
WBPk

94
 20

18

16
/WBP
BP/WBP
NBPWBP

14
of ofNBP/

Mean
12 Sigma
Sigma

10
andanSdigma

8
M ean

6
Mean

0
10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56
C/N (dB-Hz)
C/N0
0(dB-Hz)

Figure 4.11: Statistics of normalized power NP for M=20

The lock detector measurement is:

K
1
µ̂ NP =
K
∑ NP
k −1
k (4.29)

A threshold can be set at a desired minimum C/N0. If 26 dB is the minimum expected

C/N0, the average of NPk should be greater than 6. When this average value is less than 6,

it indicates a loss of lock of the DLL.

The same measurement is used to estimate C/N0 as follows (Raquet 2001):

Cˆ ⎛ 1 µˆ NP − 1 ⎞
= 10 log10 ⎜⎜ ⎟⎟ (4.30)
N0 ⎝ T M − µ̂ NP ⎠

95
Equations (4.29) and (4.30) are used in the software receiver to compute a DLL detector

value and to estimate one second average C/N0 with M=20 and K=50.

4.5 Bit Synchronization

The histogram approach is used to perform bit synchronization in this software GPS

receiver. This approach breaks a data bit period (20 ms) into 20 C/A-code 1 ms epoch

periods and senses sign changes between successive epochs. For each sensed sign

change, a corresponding histogram cell count is incremented until a count in one specific

cell exceeds the other 19 bins by a pre-specified amount. An example of one such

histogram for successful bit synchronization along with count thresholds is shown in

Figure 4.12. The procedure can be described as follows (Krumvieda et al., 2001):

30

25

NBS1 Threshold 1

20
Bin counts

15

NBS2 Threshold 2

10

0
0 2 4 6 8 10 12 14 16 18
Bin number

Figure 4.12: A successful bit synchronization histogram

96
1) A cell counter K cell is arbitrarily set and runs from 0 to 19.

2) each sensed sign change is recorded by adding 1 to the histogram cell corresponding to

K cell .

3) The process continues until one of the following occurs:

(a) Two cell counts exceed threshold NBS 2

(b) Loss of lock

(c) One cell count exceeds threshold NBS1

4) If (a) occurs, the bit synchronization fails and is reinitialized. If (b) occurs, one tries to

reestablish lock. If (c) occurs, bit synchronization is successful, and the C/A code epoch

count is reset to the correct value.

The thresholds NBS1 and NBS 2 are determined as follows:

1) The probability of making an error in determining a sign change at a desired S N 0 is

Pesc = 2 Pe (1 − Pe ) (4.32)

where

Pe = erfc' [ ]
2(S N 0 )T if a PLL is used

1 ∞
erfc' ( x ) = ∫
2
e − y 2dy
2π x

2) The number of entries, N bs , in a cell has a binominal distribution. Over Tbs seconds,

the average number of sign changes (bit transitions) is 25 Tbs , so that

NBS1 = 25 Tbs (4.33)

in the correct cell, and

97
 N bs = 50 Tbs Pesc (4.34)

in the other cells. The standard deviation of N bs in any cell is

σ N = 50Tbs Pesc (1 − Pesc )

bs
(4.35)

3) The thresholds, as well as the time interval Tbs , are selected to provide a good spread

(say 3σ) between them at a desired SNR. That is, given NBS1 , select NBS 2 and Tbs for a

desired SNR so that

25Tbs − 3 50 Tbs Pesc (1 − Pesc ) ≥ NBS 2 ≥ 50 Tbs Pesc (4.36)

4.6 Sub-frame Synchronization

Frame synchronization is required in order to process the GPS data. When timing

uncertainties are large, parity decoding is not possible since the boundaries of the words

are unknown. The following section details the procedure necessary to synchronize

frames.

The GPS navigation message frame structure is shown in Figure 4.13. Each frame is

partitioned into five sub-frames. Each sub-frame is subdivided into ten 30-bit words with

the two leading words being the telemetry (TLM) word and the handover (HOW) word.

98
 Figure 4.13: Navigation message

The structure for the TLM and HOW words is given in Figure 4.14. The TLM word is 30

bits long, occurs every six seconds in the data frame, and is the first word in each sub-

frame. Each TLM word begins with a preamble, followed by the TLM message, two

reserved bits and six parity bits.

99
 Figure 4.14: TLM word and HOW word

The HOW word is 30 bits long, occurs every six seconds in the data frame and is the

second word in the sub-frame, immediately preceded by the TLM word. The convention

is that the most significant bits (MSB) are transmitted first. The HOW word begins with

seventeen MSBs of the time-of-week (TOW) count. The full TOW count consists of the

19 least significant bits (LSB) of the 29-bit Z-count. Bit 18 is the momentum flag (for SV

configuration 000) or “alert” flag (for SV configuration 001) and bit 19 is the

synchronization flag (for SV configuration 000) or antispoof flag for (SV configuration

001) (Van Dierendonck, 1996).

The algorithm for positively finding the correct preamble and performing frame

synchronization is as follows:

100
[1] Search for either an upright or inverted preamble.

[2] When one is found, which could be a legitimate pattern somewhere else in the data

stream, a check is required to verify it is the beginning of a 30-bit word. This is

accomplished by collecting the following 22 bits and checking parity. If parity does not

pass, the candidate preamble is discarded.

[3] If parity passes, it confirms that the preamble existed at the beginning of a word. The

parity algorithm will also resolve the sign ambiguity. However, there are also other

legitimate patterns at the beginning of other words, so additional checks are required. If it

is the correct TLM word, the following word must be a HOW word that contains a

truncated Z-count. The first eight bits of this truncated Z-count can also resemble a

preamble.

[4] Parity should pass on the HOW word. If not, the frame synchronization procedure

should be restarted. Two checks can be made to verify a legitimate HOW word that the

Z-count is reasonable, and it agrees with the sub-frame count. Of course, there is a small

probability that these conditions could also occur elsewhere in the message. Thus, further

checking is required.

[5] If the HOW seems to be legitimate, provisional demodulation of the other words can

commence, and they can be stored in memory. A final check on the next preamble and

101
the next Z-count solidifies the frame synchronization. That is, the preamble is where it is

supposed to be, and the Z-count increments by one.

4.7 Raw Measurement Derivation

4.7.1 Pseudorange

A whole pseudorange measurement requires more than the DLL. The DLL provides the

high resolution or fine measure of the overall pseudorange, but the whole pseudorange

also requires some gross information. To get such gross information, certain counters

become necessary.

A pseudorange measurement can be calculated by the following equation

ρ (t ) = c[tu (t ) − t ( s ) (t − τ )] (4.37)

where

t ( s ) (t − τ ) = Z - count
+ number of navigation bits
+ number of C/A codes (4.38)
+ number of whole C/A code chips
+ fraction of C/A code chip

This situation is shown in Figure 4.15 (Misra and Enge, 2001). As shown, the arrival time

tu (t ) kept by an inner clock is defined by a transition of the receiver clock. In general,

these transitions occur sometime in the middle of a C/A code chip, and so the larger task

102
is to establish the transmission time, according to the satellite, of the received code

feature identified by the receiver clock transition. Satellite time is kept by the Z-count

which is also included in the navigation message. More specifically, the Z-count

increments every 1.5 seconds, and the navigation message starts a new sub-frame every 6

seconds. These events are used to broadcast a coarse notion of satellite time to all users,

and they are needed to compute pseudoranges.

Receive time defined

by receiver clock
Fine resolution of pseudorange within
one C/A-code is measured by
correlators
Mth C/A-code within
Nth navigation bit
First C/A-code within
Nth navigation bit ………

Z-count and transmit time

referenced to first
navigation bit in subframe
Nth navigation bit

…………………..………

One subframe (6 seconds, 300 data bits)

Figure 4.15: Pseudorange construction

Since the Z-count establishes satellite time at the beginning of each sub-frame, the

transmission time is the Z-count plus the whole number of C/A-code chips since the

beginning of the sub-frame. The elapsed time can be measured using the following

components: the whole number of navigation bits; plus the whole number of C/A-codes

since the beginning of the current navigation bits; plus the number of whole C/A-code

chips since the beginning of the current code; plus the fraction of the current chip. The

last two are measured by the DLL and the rest are measured by counters in the bit

103
synchronization and sub-frame synchronization module in the software GPS receiver

developed herein.

4.7.2 Carrier Phase and Doppler

It is much easier to obtain Doppler and carrier phase measurements than to obtain

pseudorange measurements because only the fractional part is desired. The Doppler can

be read directly from the carrier NCO while the carrier phase needs to be assisted by a

carrier counter which is used to count the integer cycles the incoming carrier has

changed. The fractional part is recorded with the carrier NCO and the summation of the

integer part and the fractional part give the carrier phase measurement since the loop is

locked.

4.8 Navigation Solution

The following work focuses on how to obtain the navigation solution after the

pseudorange and carrier phase raw measurements are derived by the software receiver.

To produce the navigation solution, C3NAVG2 TM

, a software package that has been

proven successful by the PLAN group in processing GPS and GLONASS pseudorange

data in both static and kinematic modes (Petovello, 2000) is used. It can also detect cycle

slips and can smooth pseudorange measurements with the carrier phase. C3NAVG2 TM

takes an observation file and ephemeris file as input and outputs four files containing

estimated position coordinates, estimated velocities, measure of satellite geometry and a

processing results summary.

104
 CHAPTER 5

VERIFICATION OF THE SOFTWARE GPS RECEIVER BY A REAL

IF GPS SIGNAL

5.1 Software Receiver Verification Scheme

The procedure for conducting the software GPS receiver verification begins with

collecting real IF GPS data at a known position; then feeding the data to the software

receiver; acquiring the pseudorange and Doppler measurements via the receiver tracking

loops; computing the receiver position; and, finally, comparing the computed positions

with the known position information. According Parkinson (1996), the word wide

civilian horizontal positioning error for GPS is potentially about 10 metres. Therefore, if

the position error of the software receiver is better than 10 metres, it is considered to be

performing properly. In addition, the RF GPS signal can be received and processed by a

hardware GPS receiver simultaneously. Therefore, an objective comparison can be made

between the results of the hardware GPS receiver and the results of the software GPS

receiver.

A hardware front-end, ‘GPS Signal Tap’ (Figure 5.1), made by the Accord company is

used to collect the real IF GPS signal. It is an L1 GPS receiver front-end and consists of a

105
two-stage RF down-converter. The RF signal is down-converted to the intermediate

frequency, 15.42 MHz, at the second stage. This IF signal can be sampled and stored in

binary format on a PC for analysis by the user at a programmable frequency.

Figure 5.1: Hardware front-end “GPS Signal Tap”

Eighty seconds of real IF GPS data were collected by the GPS Signal Tap at Mountain

Standard Time 10:55 AM on October 20, 2003. The antenna was fixed atop the roof of

the Calgary Centre for Innovative Technology building, University of Calgary, having

WGS84 coordinates X: -1641944.887 m, Y: -3664803.795 m, Z: 4940009.362 m. The IF

bandwidth of the Signal Tap is 2 MHz. A signal with 15.42 MHz intermediate frequency

was sampled at a sampling rate of 4.75 MHz, resulting in a base-band signal centred at

1.17 MHz. One-bit quantization was conducted. The collected data consists of 1 and 0

sequence which is stored in a binary file. For processing convenience, it was then

converted to a 1 and -1 sequence with binary format and sent to the software GPS

receiver described in Chapter 4.

106
At the same time, a Novatel OEM4 hardware GPS receiver was used to receive the RF

GPS signal; some real time processing results at a certain epoch during the signal

collecting period are shown in Figure 5.2.

Figure 5.2: OEM4 receiver tracking status

From the above figure, it can be seen that seven satellites were available. The Doppler

and C/N0 of each satellite are shown in Table 5.1.

107
 Table 5.1: Doppler and C/N0 of the observed satellites
PRN number Doppler (Hz) C/N0 (dB-Hz)
1 1453.9 49.2
2 2261.1 44.7
4 - 218.6 41.7
13 2864.3 44.0
16 1672.9 46.8
20 -1931.3 48.8
25 -2599.9 45.6

5.2 Acquisition Results Verification

The collected real IF GPS data was fed to the software GPS receiver. Acquisition was

conducted first. All 32 satellites were searched and seven satellites (1, 2, 4, 13, 16, 20,

25) were acquired. This is exactly the same as the result of the OEM4 GPS receiver. The

acquired Doppler values are shown in Table 5.2. They are similar to the Doppler values

tracked by the OEM4 receiver.

Table 5.2: Acquired Doppler values of the Signal Tap collected signal
PRN number 1 2 4 13 16 20 25
Doppler (Hz) 1466.7 2666.7 - 200 2866.7 1666.7 -1933.3 -2600

5.3 Tracking Results of the Software GPS Receiver and Navigation Solution

Verification

Data bits, Doppler, signal time delay (pseudorange), DLL lock detector, and C/N0 values

can be obtained from the tracking loops. However, only the average C/N0 and Doppler

derived from the software receiver are shown in the following figures and are compared

108
with those derived from the OEM4 GPS receiver, because not all of the other

intermediate processing results of the OEM4 receiver are available for comparison

purposes. To avoid repetition, the tracking results of the software receiver for all seven

satellites except for C/N0 and Doppler will be shown in Chapter 6 where simulator

verification is discussed.

0
0

Figure 5.3 (a): Average C/N0 and Doppler for SV1

0
0

Figure 5.3 (b): Average C/N0 and Doppler for SV2

109
 0

Figure 5.3 (c): Average C/N0 and Doppler for SV4

0
0

Figure 5.3 (d): Average C/N0 and Doppler for SV13

0
0

Figure 5.3 (e): Average C/N0 and Doppler for SV16

110
 0

Figure 5.3 (f): Average C/N0 and Doppler for SV20

0
0

Figure 5.3 (g): Average C/N0 and Doppler for SV25

For clarity of comparison, the results of the average C/N0 and Doppler from the software

receiver for each satellite and those from the OEM4 receiver results are listed in Table

5.3. The Doppler values from the software receiver were collected at GPS time 147327 s,

the same as the time of the OEM4 receiver results shown in Figure 5.1 which was

October, 20 10:55:27, Mountain Standard Time.

111
 Table 5.3: Doppler and C/N0 values obtained by the OEM4 receiver and software
GPS receiver
PRN Doppler (Hz) C/N0 (dB-Hz)
OEM4 receiver Software OEM4 receiver Software
receiver receiver
1 1453.9 1456.6 49.2 45.9
20 -1931.3 -1928.9 48.8 45.3
16 1672.9 1674.3 46.8 43.2
25 -2599.9 -2598.1 45.6 41.2
2 2261.1 2260.8 44.7 40.4
13 2864.3 2869.1 44.0 39.6
4 - 218.6 - 215.6 41.7 35.8

From Table 5.3, it can be seen that the Doppler and the average C/N0 of the software

receiver and the OEM4 receiver are very close. There is about a 3 Hz difference for the

Doppler. This may be caused by a different receiver implementation and different

oscillator frequency errors occurring between the OEM4 receiver and the Signal Tap. The

C/N0 figures estimated by the software receiver are 3.5 to 6 dB lower than that estimated

by the OEM4 receiver, but the relative value of C/N0 is identical, which is listed in

descending order in Table 5.3. This may have resulted from different C/N0 estimation

algorithms and better tracking performance of the OEM4 receiver. The comparison of

Doppler and C/N0 indicates that the software receiver can acquire and track the real IF

GPS signal.

The receiver positions are computed by C3NAVG2 TM

after the pseudorange and the

carrier phase measurements are extracted from the tracking loops as described in Chapter

4. The navigation solution of the real IF GPS signal is shown in Figure 5.4.

112
 Figure 5.4: Navigation solution of the real GPS signal

It can be seen that the navigation solution matches the known real receiver position with

an RMS of 3.8 m. A 2.5 m bias can be observed although all position points centred at

the real position. This is because the pseudorange error caused by the ionospheric error,

tropospheric error, and satellite geometry. To confirm this, it is better to compare the

navigation solution in Figure 5.4 to the solution of an OEM4 receiver. Unfortunately, the

required GPS data is not logged by OEM4 receiver simultaneously. An indirect

comparing of the results of the software receiver and OEM4 receiver is conducted as the

following description. A group of GPS data is logged by an OEM4 receiver during the

test period at another point (X: -1641889.852 m, Y: -3664879.310 m, Z: 4939969.467 m)

on the roof of the Engineering Building, University of Calgary. That point is about 100

metres away from the test point. Therefore, the navigation solution of the OEM4 receiver

for that point should include almost same effects of the ionospheric error, tropospheric

error, and satellite geometry. Figure 5.5 gives the local position of that point computed

from the raw measurements generated by OEM4 receiver.

113
 Figure 5.5: Navigation solution computed from the OEM4 receiver generated raw
measurements

Comparing Figure 5.4 and 5.5, it can be seen that the receiver positions derived by the

software receiver and OEM4 receiver have similar mean and RMS values and both of

them have a position bias in the northwest direction. This confirms that the position bias

in Figure 5.4 is caused by the ionospheric error, tropospheric error, and satellite geometry

instead of other unexpected errors resulted by the implementation of the software receiver

and the software receiver has similar navigation solution as OEM4 receiver. The software

receiver is therefore proven having ability to conduct the GPS signal processing and

positioning with a positioning accuracy less than 5 metres. In addition, it is noticed that

the standard deviation of the position derived by the software receiver is not as small as

that of the OEM4 receiver. This is because some more advanced techniques, such as

narrow correlator, are used in OEM4 receiver.

The above comparison and analysis confirms the software GPS receiver works very well.

It will be used to process the simulated IF GPS signal to check its validity in the next

chapter.

114
 CHAPTER 6

SIMULATED IF GPS SIGNAL VERIFICATION RESULTS

6.1 Simulator Verification Scheme

Verification of the simulator operation is conducted by analyzing the simulated signal

and comparing it to a real GPS signal in the time, frequency, correlation, measurement

and position domains. The real IF GPS signal collected by the Signal Tap used in Chapter

5 is simulated and stored in a data file. A Fast Fourier Transform is applied to both the

simulated and real IF GPS signals to obtain signal power spectrums so that they can be

compared in the frequency domain. Since the software GPS receiver has been verified as

described in Chapter 5, it is used to process both the Signal Tap collected data and the

simulated signals. The simulated signal is fed to the receiver, which performs acquisition

and tracking and, finally, the receiver position is computed from the raw measurements

obtained from the tracking loops. All of these results are compared with those produced

from the real IF GPS signal. The position result derived by the simulated signal is also

compared with the known receiver position. In addition, the effects of the simulation

components, such as noise, filtering, quantization and errors, on the navigation solution

are analyzed for different simulation cases.

115
6.2 Simulated Signal Description

The parameters used in signal simulation are shown in Table 6.1, which exactly copies

the parameters of the Signal Tap. In the simulation, the receiver noise, satellite clock

error, and ionospheric and tropospheric errors are simulated and filtering and quantization

are performed. The C/N0 for the signal of each satellite is simulated as a constant value

according that of the OEM4 receiver as shown in Table 5.3. Ionospheric and tropospheric

errors change a little (only tens of centimeters) during 80 seconds simulation and

different for each satellite. The simulated C/N0 and average ionospheric and tropospheric

errors are listed in Table 6.2. For the selected simulation time, the simulated signal

includes SV1, SV2, SV4, SV13, SV16, SV20, and SV25. Hereafter, the simulated signal

means it is generated with noise, filtering, 1-bit quantization, and ionospheric and

tropospheric errors added unless stated otherwise.

Table 6.1: Signal simulation parameters

Simulation time Start time: GPS time 147310, Oct.20, 2003
80 seconds
Receiver trajectory Mode 0: Static
Initial position Latitude: 51.07997674o
Longitude: - 114.13384815o
Height: 1118.596 m
C/N0 Constant C/N0
Noise density -203 dBW/Hz
Intermediate frequency 15.42 MHz
Sampling frequency 4.75 MHz
Base band frequency 1.17 MHz
Front-end Bandwidth 2 MHz

116
 Table 6.2: Simulated C/N0 and ionospheric and tropospheric errors
PRN C/N0 (dB-Hz) Ionospheric error (m) Tropospheric error (m)
1 49.2 8.8 2.6
20 48.8 11.6 3.7
16 46.8 8.1 6.8
25 45.6 8.8 4.3
2 44.7 10.8 3.3
13 44.0 10.0 2.8
4 41.7 10.1 3.8

6.3 Time Domain Verification

Because the signal was digitized, it is reduced to a 1 and -1 sequence and shows noise

behaviour in the time domain. However, some properties can be shown in the frequency

domain. The signal power spectrums of the real GPS signal and simulated signal are

shown in Figure 6.1.

(a) (b)

Figure 6.1: Power spectrum of the real and simulated IF GPS signal

117
The left-hand image in Figure 6.1 shows the power spectrum of the real IF GPS signal

collected by the Signal Tap, while the right-hand image shows that of the simulated

signal. It can be seen that the power spectrum of the real and simulated signals are of a

similar shape. However, the power spectrum amplitude of the simulated signal is a little

higher than that of the real signal. This is because of a range of external effects - such as

noise, interference, Signal Tap hardware implementation errors, as well as factors that are

presence in the real GPS signal but are not simulated by the simulator. This comparison

indicates that the simulated signal has essentially the same properties, but with a higher

signal-to-noise ratio which results in better acquisition and tracking performance as

demonstrated later.

6.4 Verification in the Correlation Domain

The alignment of the incident signal and the replica code and carrier is based on the

correlation value of the two signals. The correct alignment is identified by the

measurement of the output of the correlators. The acquisition results of both the

simulated signal and the Signal Tap collected real signal are shown in the following

figures.

118
Figure 6.2 (a): Acquisition results of the real signal and simulated signal for SV1

Figure 6.2 (b): Acquisition results of the real signal and simulated signal for SV2

Figure 6.2 (c): Acquisition results of the real signal and simulated signal for SV4

119
Figure 6.2 (d): Acquisition results of the real signal and simulated signal for SV13

Figure 6.2 (e): Acquisition results of the real signal and simulated signal for SV16

Figure 6.2 (f): Acquisition results of the real signal and simulated signal for SV20

120
 Figure 6.2 (g): Acquisition results of the real signal and simulated signal for SV25

The above acquisition results demonstrate that the simulated signal has the same acquired

Doppler as the Signal Tap collected signal. The noise floors of both signals are at the

same level, but the correlation values of the simulated signal are obviously larger than the

real signal for all satellites. The approximate correlation peak values are listed in Table

6.3.

Table 6.3: Correlation peak values of all acquired satellites

PRN Auto correlation values
Real signal Simulated signal
1 4299 4638
20 4097 4542
16 3377 3579
25 1619 2820
2 1561 2690
13 1412 2689
4 1412 1914

In the above table, the correlation values of all acquired satellites are listed in descending

order, which corresponds to the order of the C/N0 values in Table 5.3. This demonstrates,

121
once again, that the simulated signal has a higher signal-to-noise ratio than the real signal,

but the relative signal power is identical to the real signal. Therefore, the simulated signal

has similar performance as the real signal in the correlation domain.

6.5 Tracking Result Comparison

6.5.1 Data Bits

Data bit demodulation is performed by a PLL. The IPS samples are simply accumulated

for the coherent integration time, with the sign of the result determining the data bits. The

change of the sign of the IPS integration value means that the phase of the data bit is

reversed by 180o. There are usually a number of integration values in one data bit interval

depending on the coherent integration time, which is 1 ms for the software receiver in this

thesis. Figure 6.3 shows the resulting data bits of the Signal Tap collected and simulated

signal for each satellite.

Figure 6.3 (a): Data bits of the real signal and simulated signal for SV1

122
Figure 6.3 (b): Data bits of the real signal and simulated signal for SV2

Figure 6.3 (c): Data bits of the real signal and simulated signal for SV4

Figure 6.3 (d): Data bits of the real signal and simulated signal for SV13

123
Figure 6.3 (e): Data bits of the real signal and simulated signal for SV16

Figure 6.3 (f): Data bits of the real signal and simulated signal for SV20

Figure 6.3 (g): Data bits of the real signal and simulated signal for SV25

124
From above figures, it can be seen that, the amplitude of the data bits of the simulated

signal is higher than that of the real signal. It is to say the prompt coherent integration

value Ips of the simulated signal is higher. This is also because of the simulated signal has

a higher signal-to-noise ratio than that of the real signal. Another thing that can be

noticed is that the simulated signal has different data bits from the real signal; however,

the simulated data bits for all satellites are the same. This is because not all ephemeris

data are added in this batch of simulated navigation data although the simulator has the

ability to do so. Only the data bits such as the preamble, sub-frame ID, Z count, and

parity check, which greatly affect the tracking and pseudorange derivation, are available

in the above figures. Therefore, the data bits in the left-hand images are different from

those in the right-hand images.

6.5.2 Carrier- to-noise Ratio

The C/N0 estimated by the software receiver are shown in Figure 6.4.

0 0
0

Figure 6.4 (a): C/N0 of the real signal and simulated signal for SV1

125
 0 0

0
0

Figure 6.4 (b): C/N0 of the real signal and simulated signal for SV2

0 0
0

Figure 6.4 (c): C/N0 of the real signal and simulated signal for SV4

0 0
0
0

Figure 6.4 (d): C/N0 of the real signal and simulated signal for SV13

126
 0 0

0
0

Figure 6.4 (e): C/N0 of the real signal and simulated signal for SV16

0 0

0
0

Figure 6.4 (f): C/N0 of the real signal and simulated signal for SV20

0 0
0

Figure 6.4 (g): C/N0 of the real signal and simulated signal for SV25

127
It can be seen that the estimated C/N0 of the simulated signal is 1 to 2 dB higher than that

of the real signal. This confirms the results shown in Figure 6.1, 6.2, and 6.3. As

described in Chapter 3, the signal power is computed by Equation 3.19, which is the

function of the input parameters C/N0 and noise density. In the simulation, the C/N0 is set

to a constant value according the OEM4 receiver estimated value for each satellite and

the noise density is arbitrarily selected as -203 dBW/Hz. However, both the signal power

and noise density for the real signal are not known. The simulated signal power may be

close to the real value, may not exactly copy it. Therefore, the input parameters C/N0 and

noise density are main reasons of the simulated signal has a higher signal-to-noise ratio

and better performance than the real signal. This will also be shown in the following

results.

6.5.3 Delay Lock Loop Detector

The lock detector is a parameter that indicates the tracking status of the tracking loop.

The receiver can make some status changes according its value. If the lock detector value

is lower than the threshold, it means that the tracking loop has lost lock. The receiver

will, consequently, change the loop bandwidth, or even return to acquisition mode. This

function has not been developed in the software receiver as of yet, and is an excellent

candidate for development as a future recommendation. The DLL detectors for all seven

satellites are shown in the Figure 6.5.

128
 Tracking threshold Tracking threshold

Figure 6.5 (a): DLL lock detector of the real signal and simulated signal for SV1

Tracking threshold Tracking threshold

Figure 6.5 (b): DLL lock detector of the real signal and simulated signal for SV2

Tracking threshold Tracking threshold

Figure 6.5 (c): DLL lock detector of the real signal and simulated signal for SV4

129
 Tracking threshold Tracking threshold

Figure 6.5 (d): DLL lock detector of the real signal and simulated signal for SV13

Tracking threshold Tracking threshold

Figure 6.5 (e): DLL lock detector of the real signal and simulated signal for SV16

Tracking threshold Tracking threshold

Figure 6.5 (f): DLL lock detector of the real signal and simulated signal for SV20

130
 Tracking threshold Tracking threshold

Figure 6.5 (g): DLL lock detector of the real signal and simulated signal for SV25

If the DLL detector threshold is set at a desired minimum C/N0, say 26 dB-Hz, the

threshold value is 6. If the lock detector is lower than 6, the loop has lost lock. Figure 6.5

indicates that the delay lock loop of both the real and simulated signals maintain tracking.

But the loop detector value of the real signal is slightly lower and undergoes more

variation than that of the simulated signal. This is because the simulated signal has higher

C/N0 that that of the real signal as shown in Figure 6.4. The delay lock loop detector is

tightly related to the estimation of C/N0. The same measurement is used to estimate the

delay lock loop detector and C/N0 as shown in Equation 4.29 and 4.30. If the C/N0

dramatically decreased, DLL will lose lock.

6.5.4 Doppler

The Doppler is a very important feature of the signal which demonstrates the dynamic

condition of the user platform. Specifically, it is the carrier phase change rate that gives

us this information. The Doppler of all satellites are given in Figure 6.6.

131
Figure 6.6 (a): Doppler of the real signal and simulated signal for SV1

Figure 6.6 (b): Doppler of the real signal and simulated signal for SV2

Figure 6.6 (c): Doppler of the real signal and simulated signal for SV4

132
Figure 6.6 (d): Doppler of the real signal and simulated signal for SV13

Figure 6.6 (e): Doppler of the real signal and simulated signal for SV16

Figure 6.6 (f): Doppler of the real signal and simulated signal for SV20

133
 Figure 6.6 (g): Doppler of the real signal and simulated signal for SV25

It can be seen clearly that the resultant Doppler of the real signal and the simulated signal

show similar values and same trends over time; however, the former shows more

variability. The significant aspect of these comparisons is that the Doppler for the real

signal, from the beginning to around 23 s, has a slightly different changing trend than

later on. This occurs for each satellite studied, and may be the result of frequency drift of

the oscillator in the Signal Tap.

6.5.5 Pseudoranges

The pseudorange is the most important measurement for obtaining navigation solution.

Errors in the pseudorange will definitely result in the errors of the computed receiver

positions. The following figures show the pseudoranges of all satellites, as derived by the

software receiver, on a millisecond scale.

134
Figure 6.7 (a): Doppler of the real signal and simulated signal for SV1

Figure 6.7 (b): Doppler of the real signal and simulated signal for SV2

Figure 6.7 (c): Doppler of the real signal and simulated signal for SV4

135
Figure 6.7 (d): Doppler of the real signal and simulated signal for SV13

Figure 6.7 (e): Doppler of the real signal and simulated signal for SV16

Figure 6.7 (f): Doppler of the real signal and simulated signal for SV20

136
 Figure 6.7 (g): Doppler of the real signal and simulated signal for SV25

No obvious differences can be observed between the pseudoranges of the simulated and

real signals due to the comparatively coarser scale of the y-axis. But the pseudorange

changing trend indicates that the pseudorange is decreasing or increasing with satellite

motion. The changing trend is identical to the sign of the satellite Doppler. If the satellite

is moving in a direction toward the receiver, the Doppler is positive and the pseudorange

is decreasing, and vice versa.

To see clearly the difference between the pseudoranges of the simulated signal and those

of the Signal Tap collected signal, one method is to compare them both to the true range

between the satellite and the receiver. However, since it is not possible to know this

quantity, a simulated theoretical pseudorange, which consists of only the theoretical

range between the satellite and receiver and the satellite clock error computed by the

satellite clock correction parameters which are provided in the down-loaded ephemeris, is

taken as a reference. The resultant pseudoranges of both the real and simulated signals

are then compared with it. Figure 6.8 shows such pseudorange differences for all

137
satellites which are obtained by subtracting the simulated theoretical pseudoranges from

the software receiver derived pseudoranges.

(a) (b)

Figure 6.8 Pseudorange differences of the real signal and simulated signal

The values shown in the above figure demonstrate that the real and simulated signal have

different pseudorange differences. There are mainly four factors which contribute to this

result. Firstly, the simulated ionospheric and tropospheric errors are not exact simulations

of the real ones. Secondly, the multipath signals present in the real signal are not

simulated. Thirdly, the receiver clock offsets estimated by the software receiver as one of

the navigation solution parameters are different. The software receiver makes adjustment

according this receiver clock offset to the output pseudoranges for each computation

epoch. This definitely results in the different pseudoranges of the real and simulated

signals. Finally, not all errors present in the real signal are included in the simulated

signal. The figure 6.8 also shows the standard deviation of the pseudorange differences of

the simulated signal is much smaller than that of the real signal. This is because the

simulated signal has higher signal-to-noise ratio and therefore less noisy. In addition, the

138
pseudorange difference for each satellite is also different. This is because the errors, such

as ionospheric and tropospheric errors, are different for each satellite and it will result in

positioning errors as shown later.

6.6 Verification in Position Domain

The navigation solutions for both the real and simulated GPS signals processed by

C3NAVG2 TM are shown in Figures 6.9 and 6.10, respectively.

Figure 6.9: Navigation solution of the real GPS signal

Figure 6.10: Navigation solution of the simulated GPS signal

139
Comparing the above two figures, the navigation solutions of both the real and simulated

signals have a horizontal positioning accuracy better than 5 metres. The result of the

simulated signal is even better. The average receiver position derived by the simulated

signal is closer than that derived by the real signal and the standard deviation and RMS of

the receiver position in Figure 6.10 for the simulated signal are also smaller than the

results in Figure 6.9 for the real signal. This indirectly illustrates the impacts of the

simulated errors and signal-to-noise ratios. A small position bias can be observed for the

solutions of both the real and simulated signals, which results from different pseudorange

errors, such as ionopheric and tropospheric errors, for different satellites. And the bias is

different for the real and simulated signals. This is because the simulated ionospheric and

tropospheric errors are different from the real ones and some interference factors and

multipath signal which are presence in the real signal are not included in the simulated

signal. The higher signal-to-noise ratio of the simulated signal makes the smaller standard

deviation for the derived pseudorange and position.

The result for the simulated signal shown in Figure 6.10 is computed with only

pseudorange measurements. To check the validity of the carrier phase of the simulated

signal, the carrier smooth method is used in C3NAVG2 TM to calculate the navigation

solution. The carrier smoothed result is shown in Figure 6.11.

140
 Figure 6.11: Carrier smoothed navigation solution of the simulated GPS signal

Result shown in figure 6.11 shows a significant positioning accuracy improvement. The

standard deviation is greatly decreased but the mean value maintains the same as before,

which is the typical carrier smoothing behaviour. This verifies the correctness of the

simulated carrier phase from another point of the view.

Based on the above analysis, the simulated signal is confirmed to be able to produce

correct navigation solution and to have similar properties to the real signal in the position

domain.

6.7 Some Other Results for Various Simulation Cases

As described before, the simulated signal is generated by adding noise, and ionospheric

and tropospheric errors to the original signal and performing filtering, quantization. All

these components can be chosen separately in the simulation. In this section, some results

for seven different simulation cases are given to show the effects of the simulation

components. The simulation cases include:

141
 (a) Simulated signal without noise with 1-bit quantization, receiver trajectory

mode 0;

(b) Simulated signal with noise, filtering, and 1-bit quantization, receiver

trajectory mode 0;

(c) Simulated signal with noise, filtering, and 2-bit quantization, receiver

trajectory mode 0;

(d) Simulated signal with noise, ionospheric error, filtering, and 1-bit

quantization, receiver trajectory mode 0;

(e) Simulated signal with noise, ionospheric error, tropospheric error, filtering,

and 1-bit quantization, receiver trajectory mode 0;

(f) Simulated signal with noise, filtering, and 1-bit quantization, receiver

trajectory mode 1;

(g) Simulated signal with noise, filtering, and 1-bit quantization, receiver

trajectory mode 2.

For the simulated signal, the theoretical pseudorange, Doppler, and real position are

known, which will be used as the references to compare with the results derived from

different simulation cases by subtracting these theoretical values from the derived ones.

The navigation solution, pseudorange difference, and Doppler difference for all

simulation cases are given below.

142
 1-bit quantization 1-bit quantization

Figure 6.12: Navigation solution of the simulated GPS signal with 1-bit quantization

1-bit quantization 1-bit quantization

Figure 6.13: Pseudorange and Doppler differences of the simulated GPS signal
with 1-bit quantization
Case (a) is a simplest simulation case with only 1-bit quantization applied and no

simulated noise. The position RMS 0.5 m can be observed in Figure 6.12. The

pseudorange and Doppler difference results are shown in Figure 6.13. The pseudorange

difference for each satellite is tens of centimetres and the Doppler difference is several

centimetres per second (0.01 m/s corresponding to 0.05 Hz). These errors actually

include the effects of both 1-bit quantization and the tracking error of the software

receiver.

143
 Noise Noise
Filtering Filtering
1-bit quantization 1-bit quantization

Figure 6.14: Navigation solution of the simulated GPS signal

with noise, filtering and 1-bit quantization

Noise Noise
Filtering Filtering
1-bit quantization 1-bit quantization

Figure 6.15: Pseudorange and Doppler differences of the simulated GPS signal
with noise, filtering and 1-bit quantization

Figure 6.14 and 6.15 show the results for the case (b). The noise is added and the filtering

is performed comparing to the case (a). It can be seen that the RMS of the horizontal

position is greatly increased from 0.5 m to 1.9 m and standard deviation from 0.4 m to

1.9 m. The standard deviation of the pseudorange difference and Doppler difference are

increased as well. These changes demonstrate the effects of the noise and filtering. The

presence of the noise definitely reduces the signal-to-noise ratio relating to the case (a)

and causes larger tracking error and tracking error variation. Filtering limits out of band

144
noise in one hand, but filters out a part of signal power of the side lobes in the other hand.

However, ninety percent C/A code signal power is in the 2 MHz main lobe (Raquet,

2001). Therefore the application of 2 MHz filtering in this simulation limits the noise but

has no significant effect on the power of the C/A code signal.

Noise Noise
Filtering Filtering
2-bit quantization 2-bit quantization

Figure 6.16: Navigation solution of the simulated GPS signal

with noise, filtering and 2-bit quantization

Noise Noise
Filtering Filtering
2-bit quantization 2-bit quantization

Figure 6.17: Pseudorange and Doppler differences of the simulated GPS signal
with noise, filtering and 2-bit quantization

To check the effect of 2-bit quantization, a simulated signal with noise, filtering, and 2-

bit quantization is generated in case (c), the results of which are shown in Figure 6.16 and

145
6.17. Comparing with case (b), the standard deviation of position, pseudorange difference

and Doppler difference are all reduced. This demonstrates the 2-bit quantization causes

less quantization loss.

Noise Noise
Filtering Filtering
1-bit quantization 1-bit quantization
Ionospheric error Ionospheric error

Figure 6.18: Navigation solution of the simulated GPS signal

with noise, filtering, 1-bit quantization, and ionospheric error

Noise Noise
Filtering Filtering
1-bit quantization 1-bit quantization
Ionospheric error Ionospheric error

Figure 6.19: Pseudorange and Doppler differences of the simulated GPS signal
with noise, filtering, 1-bit quantization, and ionospheric error

For the case (d), the ionospheric error is added in addition to noise, filtering, and 1-bit

quantization. A position bias can be clearly seen in Figure 6.18. This is because the

different pseudorange variations of the different satellites which are resulted by different

146
simulated ionospheric errors are applied for the different satellites. Theoretically, the

pseudorange differences as shown in Figure 6.19 should be the summation of the values

in Figure 6.15 and the simulated ionospheric errors. But it is not true. This is because the

software receiver conducts adjustment to the output pseudorange according the computed

common receiver clock offset of all satellites for each computation epoch. Therefore, an

offset value should be taken into account to check the validity of the pseudorange

differences. So, the pseudorange differences for the simulated signal with ionospheric

errors as shown in Figure 6.19 should be the summation of the values for the simulated

signal without ionospheric errors as shown in Figure 6.15, the simulated ionospheric

errors, and a common offset value. But the common offset value is not known. In Table

6.4, it is computed for each satellite according other three known values. It can be seen

that the common offset value is almost the same for all satellites. This demonstrates the

relationship described above stands and the results shown in case (d) are correct.

Table 6.4 Pseudorange differences of the simulated signals

and simulated ionospheric errors
PRN Pseudorange difference Simulated Common
Simulated signal Simulated signal ionospheric offset
without ionospheric with ionospheric error (m) value (m)
error (m) error (m)
1 0.7 - 1.4 8.8 10.9
2 0.6 1.1 11.6 11.1
4 - 0.1 - 2.4 8.1 10.4
13 0.7 - 1.3 8.8 10.8
16 0.5 0.6 10.8 10.7
20 0.0 - 0.8 10.0 10.8
25 0.4 - 0.4 10.1 10.9

147
The results of case (e) have already shown in Figure 6.10. Comparing to case (d), the

only difference of the two cases is that the tropospheric error is simulated in the case (e).

However, the height biases for the two cases differ by 6.3 m. This difference is resulted

by the tropospheric error as no tropospheric model is applied when using C3NAVG2 TM to

derive the result of case (e). When the tropospheric error model is applied, this bias is

gone as shown in Figure 6.20.

Figure 6.20: Navigation solution of the simulated GPS signal with tropospheric
error model applied in C3NAVG2 TM

So far, all results and analysis are for the static receiver trajectory mode 0. The following

figures show the navigation solutions, pseudorange differences, and Doppler differences

for the receiver trajectory mode 1 and 2. Figure 6.21 and 6.22 are the results of the

simulated signal with noise, filtering, 1-bit quantization, ionospheric and tropospheric

errors for the receiver trajectory mode 1 and Figure 6.23 and 6.24 are for the receiver

trajectory mode 2.

148
 Figure 6.21: Navigation solution of the simulated GPS signal
(receiver trajectory mode 1)

Figure 6.22: Pseudorange and Doppler differences of the simulated GPS signal
(receiver trajectory mode 1)

149
 Figure 6.23: Navigation solution of the simulated GPS signal
(receiver trajectory mode 2)

Figure 6.24: Pseudorange and Doppler differences of the simulated GPS signal
(receiver trajectory mode 2)

It can be seen that the navigation solutions for both receiver trajectory mode 1 and mode

2 follow the simulated trajectory. The RMS of the positions, mean values, and standard

deviations are in the same level as the results for receiver trajectory mode 0 shown in

Figure 6.8 (b) and 6.10. This demonstrates the validity of the simulation for trajectory

mode 1 and mode 2.

150
 CHAPTER 7

CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

7.1.1 Software IF GPS Signal Simulator

A GPS signal simulator has been developed and verified in this thesis. The simulator is

based on a mathematical model at an intermediate frequency. The desired simulation

time, receiver trajectory, and other parameters such as the selected intermediate

frequency, sampling frequency, filtering bandwidth, C/N0, and noise density are the

required inputs that drive the simulator. Both filtering and quantization processing steps

were implemented. The satellite clock error, ionospheric and tropospheric errors were

simulated, with the simulator outputting a quantized bit stream. All of the above

components can be easily altered or conveniently selected or rejected in a parameter

definition file. The size of the computation burden was also optimized, by developing

simulator on MATLAB, with some heavy computational load functions developed in C

code to accelerate the simulator running speed. The simulator currently requires 100

seconds to generate 1 second of IF GPS signal.

151
The simulator has been verified by comparing the simulated signal and the Signal Tap

observed signal in the time, frequency, measurement, and position domains. Acquisition

and tracking results were also compared to prove the correctness of the simulator. It was

evident from this comparison that the simulated IF GPS signal has the same properties as

those of the hardware front-end collected real IF GPS signal in all respects. Therefore,

this confirmed that the simulator-generated signal can accurately duplicate the true signal.

Further, it demonstrates that the simulator itself can act as an effective and versatile tool

for GPS receiver studies since it is easily adapted to generate not only the GPS signals

that have already been implemented but also the signals that will be implemented in the

future, such as GPS L5 and Galileo signals.

7.1.2 Software GPS Receiver

A software GPS receiver was also developed and verified in this thesis. It is a multi-

channel receiver that operates in post-mission. Like a conventional receiver, it includes

several functional blocks: acquisition, code and carrier phase tracking, bit

synchronization, sub-frame synchronization, and navigation solution computation. This

receiver is driven by the digitized IF GPS signal. An inner clock based on the incoming

IF GPS signal samples is generated as the time reference because no physical clock is

available in the software receiver as in existing hardware receivers. The navigation data

bits, from which the satellite ephemeris can be acquired, are demodulated in the phase

lock loop. The pseudorange and carrier phase measurements can be derived from the

tracking loop with the help of bit synchronization and sub-frame synchronization.

Finally, receiver positions in the WGS84 system are output from the software receiver.

152
The software receiver has been verified by testing against real GPS signals. The real GPS

signal collected via the Signal Tap was fed to the receiver. Various tracking results,

carrier phase and pseudorange measurements, and position solution were logged to verify

the receiver correctness. For example, the Doppler derived from signal acquisition and

signal tracking indicates that it can track the carrier correctly; the resulting navigation

solutions demonstrate that the receiver can obtain satisfactory pseudorange measurements

and provide an accurate receiver position solution. From these test results, it can be safely

concluded that the software receiver operates properly and can be used for the simulated

GPS signal verification.

7.2 Future Work

7.2.1Future Work for the Software IF GPS Signal Simulator

In addition to the design principles, implementation factors, and verification processes

that have been described in this thesis, the following items are recommended for further

study:

[1] A receiver clock error model should be simulated. It is one of the major error sources

in a conventional receiver, caused by the instability of the oscillator. It has a significant

impact on maintaining coherent signal tracking.

153
[2] Multipath should be simulated. In an actual positioning environment, the received

signal is the combination of direct and multipath signals. Strong multipath signals will

deform the autocorrelation function and mislead the delay lock loop to track the wrong

autocorrelation, potentially resulting in a tracking error in the order of tens of metres.

[3] The simulator in this thesis is based on the L1 C/A code signal. It thus constitutes

only a basic application of signal simulation. New signals, such as the L5 signal and even

the Galileo signal should be simulated to study their behavior.

[4] Simulation of interference would be a valuable addition to this research. A software

method such as this one allows a great degree of flexibility in interference spectrum

analysis.

[5] Realizing the simulator using C++ would further enhance its efficiency, due to C++’s

capability to accelerate simulation speed. Developing a graphical user interface to make

the simulator easier to operate would also be valuable.

7.2.2 Future Work for the Software GPS Receiver

The software GPS receiver used in the simulated signal verification is only a basic

version because it fulfilled only the core part of a conventional receiver, such as signal

acquisition, signal tracking, raw measurement derivation, and positioning fixing. To

further improve receiver performance, the following tasks should be conducted in the

future:

154
[1] Improve the tracking accuracy. The parameters of the FLL, PLL, and DLL should be

adjusted based on automatic control and phase lock loop theory to decrease the tracking

loop errors. The parameters should be changed adaptively with respect to the receiver

status, enhancing its capability to handle either low or high dynamics, and to

accommodate light or heavy signal attenuation.

[2] Apply external aiding to the receiver to remove large dynamics or other major error

sources to improve receiver performance.

[3] Apply interference mitigation methods to improve receiver performance in hostile

environments.

[4] Upgrade the receiver to a full GNSS receiver which can process not only the GPS

C/A signal but also GPS L2, GPS L5, and Galileo signals.

155
 REFERENCES

Accord, (December 8, 2002), http://www.accord-products.com/prodgsim.htm

Akos, D. M. (1997), A Software Radio Approach To Global Navigation Satellite System

Receiver Design, Dissertation, Ohio University.

Amoroso, F. (1983), Adaptive A/D Converter to suppress CW Interference in DSPN

Spread-Spectrum Communications, IEEE 1983.

Axelrad, P., R. G. Brown (1996), Global Positioning System: Theory and Applications,

Volume I, Chapter 9: GPS Navigation Algorithms, University of Colorado 80309, Iowa

State University, Ames, Iowa 50010. Inc.: 409-433.

Brown, A., N. Gerein, and K. Taylor (2000), Modeling and simulation of GPS Using

Software Signal Generation and Digital Signal Reconstruction, Proceedings of ION

NTM, Anaheim, CA, pp.646-652.

Burns, J. R., C. Cutright, and M. Braasch (2002) Investigation of GPS Software Radio

Performance in Combating Narrow Band Interference, ION 58th Annual Meeting/CIGTF

21st Guidance Test Symposium, Albuquerque, NM, pp.523-530.

Black, H. D., A. Eisner (1984), Correcting Satellite Doppler Data for Tropospheric

Effects, Journal of Geophysical Research, Vol. 89, No. D2: 2616-2626.

156
Corbell, P. M., M. M. Miller (2000) Design and Analysis of a Matlab Based Digitized IF

GPS Signal Simulator and A Simulink Based Configurable GPS Receiver, Proceedings of

ION GPS, Salt Lake City, UT, pp. 1906-1915.

Corbell, P. M. (2000) Design and Validation of an Accurate GPS Signal And Receiver

Truth Model For Comparing Advanced Receiver Processing Techniques, Thesis for

master degree, Department of Air Force, Air University.

Corbell, P. M., M. M. Miller (2001) A Configurable GPS Accumulated I and Q Signal

Component Simulator in Matlab, Proceedings of ION NTM, Long Beach, CA, pp. 205-

215.

Christie, J. R., P. Ko, A. Hansen, D. Dai, S. Pullen, B. S. Pervan, B. W. Parkinson

(1999), The Effects of Local Ionospheric Decorrelation on LAAS: Theory and

Experimental Results, Proceedings of ION National Technical Meeting, the Institute of

Navigation, Alexandria, VA: 769-777.

CRS, (November 12, 2002), http://www.centerforremotesensing.com/press/gps.pdf

FAA (1997), Specification for the Wide Area Augmentation System, FAA-E-2892C.

Georgiadou, Y., A. Kleusberg (1988), On Carrier Signal Multipath Effects in Relative

GPS Positioning, Manuscripta Geodaetica, Vol. 13, No. 3: 172-179.

Goldhirsh, J., W. J. Vogel (1989), Mobile Satellite System Fade Statistics for Shadowing

and Multipath from Roadside Trace at UHF and L-Band, IEEE Transactions on

Antennas and Propagation, Vol. 37, No. 4: 489-498.

157
Hopfield, H. S. (1969), Two-Quartic Tropospheric Refractivity Profile for Correcting

Satellite Data, Journal of Geophysical Research, Vol. 74, No. 18: 4487-4499.

Kaplan, E. D. (1996), Understanding GPS, Principles and Applications, Boston: Artech

House, Inc.

Krumvieda K., P. Madhani et al (2001), A Complete IF Softare GPS Receiver: A Tutorial

about the Details, ION GPS 2001,11-14 September, 2001, Salt Lake City, UT.

Lachapelle, G. (1997), ENGO 625 GPS Theory and Applications. Department of

Geomatics Engineering, the University of Calgary.

Ledvina, B.M., M. L. Psiaki et al (2003), A 12-Channel Real-Time GPS L1 Software

Receiver, ION NTM.

Lin, D.M., Tsui, J. (2000), Comparison of acquisition methods for software GPS

receiver, Proceeding of ION GPS 2000, 19-22 September, 2000, Salt Lake City, UT, pp.

2385-2389.

Luo, N. (2001), Precise Relative Positioning of Multiple Moving Platforms Using GPS

Carrier Phase Observables, Ph.D. thesis, UCGE reports 20147, the University of

Calgary.

Miller, M. M., P. M. Corbell (2000) Design and Validation of Digitized Intermediate

Frequency GPS Signal and Receiver Software Models for Developing and Comparing

Advanced GPS Receiver Technologies, Proceedings of ION GPS, Salt Lake City, UT, pp.

2362-2371.

158
Maybeck, P. S. (1994), Stochastic Models, Estimation and Control, I. New York:

Academic Press, Inc. Republished, Arlington, VA: Navtech.

McDonald, K. D. and C. Hegarty (2000), Post-Modernization GPS Performance

Capabilities, ION GPS 2000 Meeting, Salt Lake City UT, 19-22 Sep 2000

Misra, P. and P. Enge (2001) Global Positioning System - Signals, Measurements, and

Performance, Ganga-Jamuna Press, Massachusetts, 2001.

NAVSYS, (November 12, 2002), http://www.navsys.com/Products/toolbox.htm

NAVTECH,(December8, 2002), http://www.navtechgps.com/supply/correlator_simulator

Parkinson, B.W. (1996), Global Positioning System: Theory and Applications, Volume I,

Chapter 11: GPS Error Analysis, Stanford University, Stanford, California 94305. Inc.:

469-483.

Petovello, M. (2000), C3NAVG2TM Operating Manual, Department of Geomatics

Engineering, The University of Calgary.

Peterson, R.L., R. E. Ziemer, D. E. Borth (1995), Introduction to Spread Spectrum

Communications, Englewood Cliffs, NJ: Prentice Hall.

Raquet, J. F. (2001), ENGO 699.64 Lecture Note: GPS Receiver Design, Department of

Geomatics Engineering, The University of Calgary.

Ray, J. K. (2000), Mitigation of GPS Code and Carrier Phase Multipath Effects using a

Multi-Antenna System, Ph.D. thesis, UCGE Report No. 20136, February 2000.

159
RTCA (1998), Minimum Operational Performance Standards for Global Positioning

System/Wide Area Augmentation System Airborne Equipment, RTCA/DO-229A, SC-159.

Washington, DC.

Schaer, S. (1997), How to Use CODE’s Global Ionosphere Maps, Astronomical Institute,

University of Berne.

Seitz, A. C., and M. S. Braasch (1998), High Fidelity GPS Receiver Simulation, ION

Proceedings of the 54th Annual Meeting, The Institute of Navigation, Colorado, pp. 573-

581.

Spilker Jr., J.J. (1996a), Global Positioning System: Theory and Applications, Volume I,

Chapter 3: GPS Signal Structure and Theoretical Performance, Stanford Telecom,

Sunnyvale, California 94089. Inc.: 57-119.

Spilker Jr., J.J. (1996b), Global Positioning System: Theory and Applications, Volume I,

Chapter 4: GPS Navigation DATA, Stanford Telecom, Sunnyvale, California 94089. Inc.:

121-176.

Spilker Jr., J.J. (1996c), Global Positioning System: Theory and Applications, Volume I,

Chapter 13: Tropospheric Effects on GPS, American Institute of Aeronautics and

Astronautics. Inc.: 517-546.

Tsui, James B-Y. (2000), Fundamentals of Global Positioning System Receivers: A

Software Approach, John Wiley & Sons Inc.

160
Van Dierendonck, A.J. (1996), Global Positioning System: Theory and Applications,

Volume I, Chapter 8: GPS Receivers, AJ Systems, Los Altos, CA 94024. Inc.: 329-407.

Van Vee, R.D.J., Coenen. A.J.R.M. (1991), New Fast GPS Code-Acquisition Technique

Using FFT, Electronic Letters, Vol. 27, pp.158-160, Jan. 1991.

161